Quantum Computing Business Model: Where the Money Is

How Will Companies Actually Make Money from Quantum Technology?

◆

On 21 May 2026, the United States took a step that made one thing clear: Quantum Technology is no longer merely a scientific experiment; it is on its way to becoming a multibillion-dollar business. The US Department of Commerce entered into agreements to invest a total of $2.013 billion in nine quantum companies. Its biggest bet was on IBM, which was set to receive up to roughly $1 billion to establish a separate quantum foundry in Albany, a facility where the quantum chips and hardware of the future could be developed and manufactured. GlobalFoundries was allotted up to $375 million, while Atom Computing, D-Wave, Infleqtion, PsiQuantum, Quantinuum and Rigetti were each in line for up to $100 million. Up to $38 million was set aside for Diraq. But the most interesting part was that this money was not being handed out merely as a subsidy. In return, the US government would receive small, non-controlling stakes in these companies. Put simply, America is not only supporting Quantum Technology; it is also betting on a share of its future profits.

Now consider another story from the same period.

In 2025, quantum company D-Wave published a paper in Science and claimed that its Advantage2 quantum annealer had simulated the dynamics of an extremely complex physical system called a spin glass with a degree of accuracy that no classical computer could achieve. The company’s CEO went so far as to call it “quantum supremacy.” But about a year later, the story appeared to turn. Researchers at the Flatiron Institute and Boston University published a new paper in Science in which they combined modern tensor-network techniques with an old belief-propagation method from 1982 to create a classical algorithm capable of reproducing many of D-Wave’s benchmark calculations on ordinary computers. Some calculations ran on a workstation and, according to some reports, even on a personal laptop. D-Wave disputed the conclusion, arguing that the largest and most difficult problems were still beyond the reach of classical computers, so the debate was not over.

When these two stories are placed side by side, an uncomfortable question emerges.

And this is where the real question begins: if much of a task described today as “impossible for classical computers” can be done tomorrow by a better algorithm running on an ordinary computer, why are governments and companies pouring billions of dollars into Quantum Technology? What future are IBM, Google, Microsoft, Amazon, pharmaceutical companies, banks, defence agencies and venture capital firms trying to buy?

Where, exactly, will the money be made?

This article follows that question. But the answer will not be found where most people look for it. The answer is not inside the quantum computer. It lies in the entire industrial structure being built around the quantum computer, and a large part of that structure is not connected to quantum computing at all.

The First and Most Important Point: Quantum Physics Is Not Quantum Computing

This is the most important sentence in the article, so it is better to state it at the outset.

Quantum computing is only one of the many commercial branches to emerge from quantum physics. It is the most talked-about branch and the one attracting the most money, but it is neither the oldest nor the most profitable, and it may not even be the first to mature.

On one side, consider the technologies that could not exist without quantum mechanics: the transistor, laser, LED, solar cell, MRI, semiconductor memory, fibre-optic communication, electron microscope and nuclear medicine. All are direct descendants of quantum physics. Together, they are called the “first quantum revolution,” and they generate trillions of dollars in annual business. Almost the entire electronics industry rests on this foundation.

But if we use that fact to call every electronics company a “quantum company,” the term loses all commercial meaning. Intel is not a quantum company simply because a transistor depends on quantum tunnelling. Samsung is not a quantum company simply because electrons cross barriers inside its flash memory. If everything is quantum, the word quantum tells us nothing.

So what does today’s “quantum industry” actually mean?

From both a commercial and a technical perspective, the answer is this: the second quantum revolution begins when we learn to create, control and measure individual quantum states, and to make them perform useful work.

In the first revolution, we benefited from quantum rules collectively. Millions of electrons work together to form a transistor, and we do not care about any single electron. In the second revolution, we isolate one atom, one ion, one photon or one superconducting circuit and communicate with it. We place it in superposition, entangle it with other particles, and preserve its fragile state long enough to perform a calculation or measurement.

This distinction is not merely academic. It is the line that determines whether a company is genuinely a quantum business or is simply using the word quantum in its marketing.

Under this definition, today’s quantum technology economy contains at least thirteen distinct commercial fields:

1. Quantum Computing, meaning general-purpose computation using gate-model and other architectures
2. Quantum Annealing, a separate approach designed for a particular class of optimisation
3. Quantum Simulation, in which one quantum system is used to understand another
4. Quantum Communication, transmitting information through entanglement and photons
5. Quantum Cryptography, especially Quantum Key Distribution
6. Post-Quantum Cryptography, which has quantum in its name but runs on classical computers
7. Quantum Sensing, using atoms and spins to make extraordinarily fine measurements
8. Quantum Metrology, the science of measurement units and standards
9. Quantum Imaging, where it makes commercial sense
10. Atomic Clocks and precision timing
11. GPS-independent navigation
12. Quantum Materials, including superconductors, topological materials and engineered lattices
13. Enabling infrastructure: cryogenics, lasers, photonics, control electronics, detectors and fabrication

Some of these fields are already profitable. Some will make money in the next decade. Some may never become large businesses. The purpose of this article is to explain which is which, and why.

Before discussing the business, we need a little foundational science, but only enough to understand the commercial logic. Read this not as a textbook, but as a map that will help with every economic question that follows.

Bits and Qubits

A classical bit is in one of two states: 0 or 1. It is like a switch, either on or off.

A qubit can be in a mathematical combination of both states, which is called superposition. It is written as a weighted combination of 0 and 1, where the weights are complex numbers called amplitudes.

This is where the first misconception arises, and it has found its way into the marketing of the entire quantum industry.

People say: “A quantum computer tries every answer at once.”

That statement is not technically false, but it hides the part that matters most. Yes, the state of n qubits is described by 2 to the power of n amplitudes. Describing the state of 300 qubits would require more numbers than there are atoms in the universe. That is true.

But when you measure the system, you receive only one n-bit answer. One. Everything else disappears.

So a “parallel universe” containing 2 to the power of n answers is useless if you can extract only one random answer from it. You cannot place a 300-qubit register in superposition and read 2^300 answers from it. You can read one answer, and even that arrives according to probability.

The Real Magic Is Not Superposition, but Interference

The real work of a quantum algorithm is to rotate the amplitudes in such a way that the amplitudes of the wrong answers cancel one another out while the amplitude of the right answer grows. When the system is measured at the end, the probability of receiving the correct answer has become very high.

The clearest analogy is a pair of noise-cancelling headphones. They generate an inverted version of the external sound wave. When the two waves meet, they cancel one another, and you hear silence. A quantum algorithm does something similar, not with sound waves but with probability amplitudes.

Where the analogy breaks down: sound waves physically exist in the air and can be measured with a microphone. Quantum amplitudes do not exist in physical space. They are complex numbers whose squared magnitudes give us probabilities, and they cannot be measured directly. The headphone picture is therefore useful for intuition but incomplete as physics.

The direct commercial conclusion is this: a quantum computer cannot accelerate every problem. It can accelerate only those problems for which someone has discovered an interference pattern that amplifies the correct answer. There are very few such patterns. Even after thirty years of research, the list of commercially valuable quantum algorithms remains short, and most of them require hardware that does not yet exist.

Entanglement

When two or more qubits are entangled, their states cannot be described independently. The group has one shared state. Measuring one immediately reveals information about the other, no matter how far apart they are.

Here, too, there is a popular misconception. Entanglement cannot transmit information faster than light. The person making the first measurement receives a random result and cannot choose it. The second person still needs a message through an ordinary classical channel to know how their result corresponds to the first. That is why quantum communication is not an “instant internet,” and why the commercial case for a quantum internet rests not on being “faster,” but on being “different.” We will return to this in detail.

Decoherence: The Biggest Cost in the Entire Industry

A qubit’s superposition is extremely fragile. Ambient temperature, electromagnetic noise, a molecule of air, even a cosmic ray can effectively “measure” it and destroy its quantum state. This is called decoherence.

This one word explains almost the entire cost structure of the quantum industry.

Superconducting qubits must be kept below 10 millikelvin, hundreds of times colder than the average temperature of space. That requires a dilution refrigerator. Trapped-ion computers require an ultra-high vacuum and exceptionally stable lasers. Neutral atoms must be held with optical tweezers. Photonic systems need detection at the single-photon level.

Together, these facts tell us one thing: quantum computing is not a software business. It is a heavy, expensive and slow hardware business. That fact lies at the root of every economic question in the industry.

Physical Qubits and Logical Qubits: The Distinction That Creates the Most Confusion

Because qubits continually suffer errors, we need quantum error correction.

Error correction is easy in classical computers. You make three copies of a bit, and if one changes, the majority tells you the correct value. That is not possible in quantum computing because the no-cloning theorem says that an unknown quantum state cannot be copied exactly.

Quantum error correction therefore uses a clever method: it spreads information across several physical qubits in such a way that an error can be detected without reading the state itself. Several physical qubits combine to create a more reliable logical qubit.

The threshold theorem says that if the error rate of the physical qubits is below a certain limit, increasing the size of the code can reduce the logical error rate exponentially. In December 2024, Google demonstrated on its Willow chip, for the first time, that the logical error rate did indeed fall as the size of the surface code increased. It was a decisive moment for the field.

Now for the point every reader should remember, because it places every press release in the correct perspective:

A machine containing thousands of physical qubits does not give a customer thousands of reliable computational qubits.

Several verified figures make the distinction clear. In November 2025, Quantinuum used 98 physical barium ions on its Helios trapped-ion system to create 48 logical qubits, a ratio of roughly two to one and the most efficient encoding achieved so far. In January 2026, QuEra reported in Nature that it had used 448 neutral atoms to demonstrate 96 logical qubits with a high-rate code. By contrast, conventional estimates based on surface codes suggest that a single good logical qubit may require a thousand or more physical qubits.

IBM’s target illustrates the difference most clearly. According to the company’s public roadmap, its Starling system, planned for 2029, will use roughly ten thousand physical qubits to create 200 logical qubits and run one hundred million gate operations. Ten thousand in exchange for two hundred. That is the real exchange rate of quantum computing.

Why More Qubits Do Not Necessarily Mean a Better Computer

If someone tells you only the qubit count and nothing else, they have told you nothing.

A useful quantum computer needs at least five things at the same time: enough qubits, extremely high gate fidelity, good connectivity between qubits, long coherence times and fast readout. A weakness in any one of them can render all the others useless.

The clearest evidence comes from IBM’s own product strategy. In 2023, the company built Condor with 1,121 qubits, but it did not make it its main commercial product. Instead, IBM focused on Heron, which had fewer qubits. Nighthawk, announced in November 2025, has 120 qubits, fewer than Heron’s 133, but its square lattice and 218 tunable couplers allow it to run circuits that are roughly 30 per cent more complex than before.

Increasing the number of qubits is comparatively easy. Keeping them all good at the same time is difficult. Commercial value comes from the second achievement, not the first.

The Quantum Business Stack: Seven Layers, Seven Different Businesses

Now we come to the real question. Where does the money come from?

The greatest analytical mistake is to treat the quantum industry as if it were a single company. Think of it instead as a layered structure, in which every layer has its own product, customer, revenue model, risk and timeline.

Layer 1: Fundamental Research

At the bottom are universities, national laboratories and government research institutions.

Who pays: governments, science ministries, defence research agencies and, to a smaller extent, the R&D divisions of large corporations.

Why they pay: because the knowledge created here does not appear on any one company’s balance sheet, but none of the layers above can exist without it. Error-correction codes, qubit architectures and laser-cooling techniques all originate here.

What emerges from this layer: publications, patents, trained scientists and, most importantly, spin-off companies. PsiQuantum, QuEra, Pasqal, Quantinuum, IonQ, Rigetti, IQM and Quandela, like almost every major quantum company, emerged from one university laboratory or another.

This layer does not earn profit directly. It is the ground on which everything else stands.

Layer 2: Physical Components, the Real “Picks and Shovels”

During a gold rush, the most dependable money was often made not by the people digging for gold, but by those selling the shovels. The saying is unusually accurate for the quantum industry, although it has limits of its own, which we will examine later.

What does it take to run a quantum computer?

Cryogenic systems. Superconducting qubits require dilution refrigerators capable of maintaining temperatures below 10 millikelvin. Finland’s Bluefors and Britain’s Oxford Instruments are the largest names in this market, alongside companies such as Leiden Cryogenics in the Netherlands, Kiutra in Germany and Maybell Quantum in the United States. In May 2025, Bluefors supplied 18 KIDE cryogenic platform systems to the G-QuAT centre at AIST Tsukuba in Japan, each capable of supporting more than one thousand qubits. The company has also acquired US cryocooler manufacturer Cryomech.

Wiring and interconnects. This is a less discussed but serious industry bottleneck. Every qubit requires several control and readout lines, so a thousand-qubit system may need three to five thousand separate cryogenic connections. It is called the “wiring crisis,” and it is driving the development of cryogenic CMOS, single-flux-quantum electronics and optical control.

Lasers and photonics. Trapped ions and neutral atoms need exceptionally stable lasers to trap, cool and control them. TOPTICA Photonics, M Squared and similar companies manufacture these systems.

Control electronics. Microwave pulse generators, arbitrary waveform generators and FPGA-based controllers. Quantum Machines, Qblox and Zurich Instruments are leading companies in this category.

Detectors. Superconducting nanowire single-photon detectors are essential for both photonic quantum computing and quantum communication.

Special materials. Isotopically enriched silicon-28 for spin qubits, quantum-grade CVD diamond for NV-centre sensors, and helium-3 for dilution refrigerators. The global supply of helium-3 is so limited that in 2026 Bluefors signed an agreement with lunar-mining startup Interlune for the supply of up to ten thousand litres a year.

Why this layer makes money today: because it is independent of the question of which qubit architecture will win. Whether the winner is IBM, IonQ or PsiQuantum, someone will still have to buy cryostats, lasers and control electronics. Every government laboratory, university and startup is a customer.

Revenue model: conventional industrial equipment sales. A one-time capital sale is followed by annual service contracts, spare parts, calibration and upgrades. The margins resemble precision manufacturing rather than software. But the revenue is dependable, and it exists today.

Risk: the customer base is small. If quantum funding falls sharply, orders will fall immediately. In addition, if one architecture wins, demand for the components it needs will rise while demand for the others will decline.

Layer 3: Quantum Hardware, Where No Winner Has Yet Been Chosen

This layer attracts the most capital and contains the greatest uncertainty. It is important to understand that science has not yet decided which path is the right one. Anyone who tells you they know the winner is either an investor or a seller.

Superconducting qubits. Metal circuits cooled to millikelvin temperatures and controlled with microwave pulses. IBM, Google, Rigetti, IQM and several Chinese laboratories follow this route.

Strength: they can be built with semiconductor-fabrication tools, their gates are extremely fast, operating in nanoseconds, and this is the most mature and heavily funded approach.

Weakness: coherence times are relatively short, every qubit requires separate wiring, and the cryogenic infrastructure is extremely expensive.

The real bottleneck: wiring and scaling. A million qubits cannot be served by a million individual wires.

Commercial strategy: IBM has chosen cloud access and enterprise partnerships, and keeps its roadmap public, which is itself a marketing weapon. In May 2026, IBM announced that it would combine $1 billion from the US government with $1 billion of its own to establish a separate quantum foundry company in Albany called Anderon.

Trapped ions. Individual atoms are ionised, suspended in electromagnetic traps and controlled with lasers. IonQ, Quantinuum, AQT and Universal Quantum follow this route.

Strength: the highest gate fidelity and the best connectivity. Two-qubit gate fidelity on Quantinuum’s Helios has been reported at about 99.92 per cent. Every ion is made by nature and is therefore identical, while manufactured qubits can vary.

Weakness: gates are much slower, operating in microseconds, thousands of times slower than superconducting gates.

The real bottleneck: scaling. Holding thousands of ions in one trap is difficult, which is why companies are working on modular trap architectures.

Commercial strategy: over eighteen months, IonQ made acquisitions worth roughly $2.5 billion, including Oxford Ionics, ID Quantique, Capella Space, Qubitekk, Lightsynq and Vector Atomic, transforming itself into a full-stack platform spanning computing, networking, sensing and space. Quantinuum went public in June 2026, selling 28 million shares at $60 each, raising roughly $1.68 billion and receiving an initial valuation of about $15.6 billion.

Neutral atoms. Neutral atoms held in optical tweezers and entangled through Rydberg states. QuEra, Pasqal, Atom Computing, Infleqtion and planqc belong to this category.

Strength: large numbers of atoms can be arranged comparatively easily, and the atoms can be physically moved to alter their connectivity, which is extremely useful for error correction.

Weakness: atoms escape from the traps, gates operate at moderate speeds, and the technology is relatively new.

The real bottleneck: continuous operation without interruption and fast error-correction cycles.

Recent progress: in July 2025, QuEra, Harvard and MIT published in Nature the first experimental demonstration of magic-state distillation at the logical level, a long-awaited step towards universal fault-tolerant computing.

Photonic quantum computing. Using particles of light as qubits. PsiQuantum, Xanadu, Quandela, ORCA and QuiX are pursuing this approach.

Strength: photons maintain coherence at room temperature and can travel through fibre, making networking natural. Existing silicon-photonics fabs can also be used.

Weakness: photons do not interact directly with one another, so gates are measurement-based and probabilistic. Detectors still need cooling, although at a less demanding level than a full cryostat.

Commercial strategy: PsiQuantum has bypassed the NISQ era entirely and is betting directly on building a million-qubit fault-tolerant machine using GlobalFoundries’ silicon-photonics manufacturing line. In May 2026, the company raised $1.5 billion at a valuation of $10.5 billion. It is a bold and risky strategy because there is no revenue along the way.

Semiconductor spin qubits. Using the spin of individual electrons in silicon as qubits. Intel, Diraq, Quantum Motion and Quobly are working on this approach.

Strength: the smallest physical footprint and, in principle, compatibility with existing CMOS fabs, which offers the most credible path to scale.

Weakness: qubit counts are currently the lowest, and uniformity is a serious problem.

Economic logic: if it works, this is the only approach that connects directly to the scale and economics of the existing semiconductor industry.

Topological qubits. Microsoft’s bet, in which quantum information is stored in the topological properties of matter so that it is naturally protected from errors. In February 2025, the company announced its Majorana 1 chip. If the approach works, it could dramatically reduce the burden of error correction. But it is also the most disputed claim in the field, and independent verification remains an open question.

Quantum annealing. This is a separate category and will be examined in detail later.

Bosonic and cat qubits. Companies such as Alice & Bob and Nord Quantique are trying to build error protection within a single physical mode, which could reduce overhead.

The commercial summary of this entire layer is this: none of these companies is profitable today, and almost all of them earn revenue that is small relative to their expenditure. These companies currently depend on capital markets, government contracts and strategic investors, not customers.

Layer 4: Control, Compilers and Middleware

Between hardware and applications lies an entire world that most articles ignore, even though a genuine software business may emerge here.

It includes pulse-level control software; qubit calibration and characterisation tools; quantum compilers that translate an abstract circuit onto the physical layout of a specific chip; error-mitigation techniques; real-time decoders that read measurement data during error correction and decide where corrections are needed; benchmarking tools; and hybrid orchestration that allocates work between quantum and classical resources.

The scale of the decoding problem is important. Every error-correction cycle produces a stream of measurement data, and it must be processed within microseconds or errors begin to accumulate. In November 2025, IBM said it had achieved real-time decoding of qLDPC codes in less than 480 nanoseconds, about ten times faster than existing methods. In March 2026, QuEra integrated with NVIDIA’s NVQLink architecture to increase decoding throughput.

This may sound technical, but its economic meaning is substantial. Error correction is also a classical computing problem. Whichever company develops the best decoder can sell it to every quantum-hardware manufacturer. For GPU companies, this is a new market.

Who pays: hardware companies themselves, national laboratories and quantum-cloud providers. Companies such as Q-CTRL have turned error-suppression software into an independent product licensed by hardware manufacturers.

Why this layer is attractive: it can approach software margins, remain relatively independent of hardware architecture and create switching costs.

Layer 5: Quantum Cloud, or Quantum Computing as a Service

This is where almost all direct customer revenue in quantum computing is generated today.

The commercial structure is straightforward: a hardware manufacturer places its machine in a data centre, a cloud provider sells access through an API, a customer submits a circuit from anywhere, and payment is based on usage.

AWS Braket, Microsoft Azure Quantum, IBM Quantum Platform and Google Cloud are the principal distribution channels in this layer.

Now consider the actual prices, because this degree of clarity is difficult to find elsewhere. According to AWS Braket’s public price list:

Here, a “shot” means running a circuit once and measuring it. Because quantum computation is probabilistic, a circuit has to be run thousands of times to obtain a meaningful answer.

Now translate these numbers into commercial terms.

Suppose a research team is running a variational algorithm that requires a total of two hundred thousand shots. On IonQ Forte, that would cost roughly $16,000. On Rigetti Ankaa, the same two hundred thousand shots would cost about $180. That is a difference of almost ninety times, reflecting the trade-off between fidelity and speed. Higher-quality hardware is expensive; cheaper hardware is noisy and therefore requires more shots, which takes back part of the saving.

Now ask the question at the heart of this entire article: how much would the same calculation cost on a classical laptop?

In most cases today, a few rupees. The reason is that a good classical simulator can also run most circuits that current hardware is capable of running.

That is the most uncomfortable truth about the quantum-cloud business. Customers today do not buy quantum time because they need the answer. They buy it because they want to learn, experiment, prepare their teams and position themselves for the future.

Where the comparison with cloud computing works, and where it breaks down.

The similarity is that both models share an expensive, specialised machine among many customers, so a customer does not need to buy it outright.

The difference appears in three places, and all three are serious. First, the server AWS sold in 2006 could solve a customer’s problem that day. Today’s QPU cannot solve most commercial problems. Second, classical cloud infrastructure can maintain high utilisation because demand is continuous. QPUs often sit idle. Third, cloud economics rested on hardware costs falling quickly while usage rose quickly. The cost of quantum hardware is not falling at the same pace.

Layer 6: Quantum Software

This layer includes algorithm-development frameworks such as Qiskit, Cirq, PennyLane and Bloqade; application-specific software; chemistry and optimisation platforms; and industry-specific tools. Algorithmiq, Multiverse Computing, QC Ware, Qubit Pharmaceuticals, ProteinQure and Kvantify are among the companies in this layer.

Here we must ask a difficult question, one that people inside the industry ask in private but seldom on stage.

Were quantum-software companies created too early?

The argument in their favour says no. Finding algorithms takes decades, and if no algorithms exist when the hardware is ready, the hardware will sit useless. These companies have also produced a genuine by-product called “quantum-inspired” classical algorithms. These are classical methods inspired by quantum thinking, and they deliver real value to real customers today without using a single qubit. A large share of the present revenue of many quantum-software companies comes from exactly this work.

The argument against them says yes, and identifies a serious structural problem. The algorithm written today for noisy 100-qubit hardware may be irrelevant on fault-tolerant hardware in 2030, because the rules of algorithm design change completely on an error-corrected machine. In the meantime, the company must pay its employees for ten years. And if its true product is quantum-inspired classical software, then it is a classical optimisation company with quantum in its name, competing against McKinsey, Gurobi and dozens of established players in the same market.

The truth is probably somewhere in the middle. Companies that generate real classical value today and use that income to build quantum capability may survive. Those waiting only for future hardware may run out of capital before the customers arrive.

Layer 7: Industry Solutions

This is the final layer, where genuine economic value is either created or not created. It is the largest part of this article, and we now turn to it.

The NISQ Era, and a Term That Is Now Growing Old

In 2018, physicist John Preskill gave today’s machines a name: NISQ, or Noisy Intermediate-Scale Quantum. It described machines containing enough qubits to make classical simulation difficult, but so much noise that error correction remained impossible.

The term remained accurate for nearly a decade. It is now becoming blurred because, between 2024 and 2026, several machines demonstrated real error correction at a small scale. We are no longer fully in the NISQ era, but neither have we entered the fault-tolerant era. We are in a transition period that does not yet have a good name.

The commercial strategy of the NISQ era was to learn to live with noise. Three tools emerged. The first was error mitigation, in which errors are not removed; instead, a mathematically corrected answer is estimated from several noisy results. It works, but its cost rises exponentially with circuit size and therefore does not scale. The second was hybrid algorithms such as VQE and QAOA, in which a quantum computer performs one small task and a classical computer handles everything else. The third was keeping the circuit short, so that the work finishes before noise accumulates.

All three share the same weakness: they operate in the very range where classical computers are also quite capable. That is why, despite a decade of effort, no commercial application has been found on NISQ machines for which a customer voluntarily pays again and again.

Fault Tolerance Changes the Entire Economic Equation

If error-corrected machines arrive, the commercial structure changes for three reasons.

First, long circuits become possible. Shor’s algorithm, quantum phase estimation and accurate chemistry simulation require millions or billions of operations. They will never run on a noisy machine, no matter how many qubits it contains.

Second, the results become dependable. No enterprise will base decisions on an answer whose accuracy cannot be guaranteed. A pharmaceutical company will commit millions of rupees on the strength of a molecular calculation only if it trusts that calculation.

Third, pricing changes. Today’s pricing is based on shots because every shot is noisy. In the fault-tolerant era, pricing will be based on logical-qubit hours or completed calculations, which is much closer to present-day cloud-computing models.

And here, remember once again the exchange rate that determines the industry’s timeline: ten thousand physical qubits for two hundred logical qubits. A complete simulation of a pharmaceutically relevant molecule needs hundreds to thousands of logical qubits, implying millions of physical qubits. The largest verified achievement today is 96 logical qubits.

That is not a small gap. But neither is it regarded as impossible to close. The valuation of the entire industry rests on this uncertainty.

Four Different Meanings of Quantum Advantage That Are Constantly Mixed Together

This is the industry’s greatest abuse of terminology. Often it is not deliberate deception, but carelessness.

Quantum Supremacy means that a quantum computer has performed a task that a classical computer cannot complete in a practical amount of time. The task does not have to be useful. Google’s 2019 random-circuit-sampling experiment belonged to this category. Better classical algorithms later weakened the claim substantially.

Quantum Advantage means that a quantum computer performed a task better or faster than a classical computer, and that the task had at least some scientific meaning. In October 2025, Google ran an algorithm called “Quantum Echoes” on its 105-qubit Willow chip to measure out-of-time-order correlators. In work published in Nature, the company claimed it was roughly thirteen thousand times faster than the best-known classical method. Importantly, the result was verifiable, meaning it could be repeated and checked. This was qualitatively stronger than the 2019 claim.

Practical Quantum Advantage means that a quantum computer solved a real industrial or scientific problem better, one that someone already wanted to solve.

Commercial Quantum Advantage means that a customer voluntarily paid the market rate for the work, repeatedly, because the quantum route was cheaper or better than the classical route.

We have reached or are approaching the first three. No public, verified example of the fourth exists yet.

One question will keep returning throughout this article, and it is the question that separates science from business:

Would a customer pay for this calculation?

Apply that question to Google’s Quantum Echoes experiment. It is a remarkable scientific achievement. Measuring OTOCs matters in quantum many-body physics. Working with Berkeley, the team applied the method to NMR data to demonstrate a “molecular ruler” that reveals information about molecular structure and could potentially be useful in drug discovery and materials science. But no pharmaceutical company is currently writing a cheque for this service every month. It is a scientific achievement, not yet a commercial product. Saying so does not diminish Google’s achievement; it simply puts it in the correct category.

D-Wave and Quantum Annealing: A Different Route, a Different Debate

Readers should clearly understand that gate-model quantum computing and quantum annealing are not the same thing.

A gate-model machine is a general-purpose computer. In principle, it can run any quantum algorithm, including Shor’s algorithm.

A quantum annealer is a special-purpose device. It performs one task: finding the lowest point in an energy landscape. Imagine a mountainous region in which you must locate the deepest valley. The annealer encodes that entire landscape into a physical system and slowly cools it towards its lowest-energy state.

Many real problems can be written in this form: selecting a portfolio, routing vehicles, assigning employee shifts and scheduling a factory. That is why D-Wave began building commercial customers much earlier than gate-model companies and today has an Advantage2 system with more than five thousand qubits.

Commercial position: D-Wave presents itself as quantum computing that is “available today,” arguing that gate-model machines are still years away. The company sells cloud access, on-premise systems and hybrid solvers. In May 2026, it received a $100 million equity investment from the US Department of Commerce.

The debate: the scientific community is deeply divided over annealing, and the disagreement is genuine. In March 2025, D-Wave published its beyond-classical claim about spin-glass dynamics in Science. Almost immediately, a team at EPFL used time-dependent variational Monte Carlo, while a team at the Flatiron Institute used belief propagation and tensor networks, to show that many of the calculations could be performed classically. Flatiron’s expanded work appeared in Science in 2026 and said that much of the result could be reproduced on ordinary workstations. D-Wave issued a formal response arguing that the most complex lattice geometries, the largest three-dimensional scales and the highest-order physical observables remained beyond the reach of classical methods.

For a commercial reader, the essence of the debate is this: in practical optimisation, it has not yet been demonstrated that annealing consistently outperforms classical solvers. In many published comparisons, a well-tuned classical heuristic performs as well or better. D-Wave’s customers often use hybrid solvers in which the classical component does most of the work.

This is not a criticism of D-Wave. It is a reminder that a technology being available and a technology being better are two different claims.

The Real Economic Question Behind Quantum Advantage

This is the most important section of the article, and, surprisingly, almost every popular account leaves it out.

Suppose a quantum computer genuinely performs a calculation faster than a classical computer. Should a company buy it?

Not necessarily. Business does not purchase speed. It purchases outcomes per rupee.

Consider a hypothetical but realistic comparison. The figures below are invented solely for illustration and are not taken from a real contract:

The quantum route is ten times faster and two hundred times more expensive. For any finance department, this is not a difficult decision.

The factors that are actually weighed in a commercial decision are these:

Time-to-solution means total elapsed time, not only machine runtime. It includes waiting in a queue, preparing data, compiling the circuit, running it repeatedly and verifying the result. A calculation that takes one second on the machine but three days from beginning to end is, in practical terms, a three-day calculation.

Cost-to-solution includes more than the cloud bill. It includes the salary of the specialist who can write the circuit, and such people are rare and expensive.

Accuracy and reliability. A wrong answer is expensive even when it is free.

Energy. This is a real and under-discussed argument in favour of quantum. A dilution refrigerator uses roughly 25 kW whether it contains 50 qubits or 5,000, because most of the energy goes into cooling. A large classical simulation may require thousands of GPUs. If a fault-tolerant machine performs the same work, the energy saving could be dramatic. It is an argument for the future, but a serious one.

Integration cost. Adding a new computational tool to a large organisation is not only a technical exercise. It involves data-security review, vendor assessment, compliance checks and internal training. In many organisations, this process takes eighteen months.

Opportunity cost. The same money and the same engineers might produce a higher return if invested in a GPU cluster or a better classical algorithm. This is the most real competitor to a quantum project.

Classical Competition: The Target That Keeps Moving

This is the point that quantum-industry marketing hides most often.

Quantum computers are not competing against the classical computers of today. They are competing against whichever classical computer exists on the day the quantum machine is ready.

And that classical computer is not standing still. It is running.

GPU capability improves with each generation. AI accelerators have become a new category. Tensor-network methods have improved so dramatically that they have weakened one “quantum supremacy” claim after another. Neural quantum states, which use machine learning to simulate quantum systems, are a rapidly growing field. Most importantly, every time a quantum claim is made, classical-algorithm researchers around the world begin trying to break it because the problem is academically attractive.

The D-Wave versus Flatiron story is the clearest example. The same happened with Google’s 2019 claim. The pattern is now so regular that it should be treated as a structural feature of the industry, not an accident.

The investment implication is serious. A quantum company may fail not only because its technology does not work, but because by the time it is ready, the classical world has already solved the problem.

There is an irony here: this competition is good for science. Every classical rebuttal defines the boundary of quantum advantage more precisely. The day a quantum result appears that classical-algorithm experts around the world cannot break, it will be far more credible precisely because so many people tried.

Part Two: Industry by Industry, Where the Money Is

Each industry below is examined through the same framework: what the problem is, how it is solved today, what quantum can do differently, where economic value is created, who writes the cheque, how a company earns revenue, how mature the technology is today, and what the biggest obstacle is.

1. The Pharmaceutical Industry and Drug Discovery

The problem. Bringing a new drug to market takes, on average, more than a decade and billions of dollars, and most candidate molecules fail along the way. Part of that failure comes from our inability to know precisely how a molecule will bind to a protein in the human body.

Today’s solution. Computational chemistry, from Density Functional Theory to coupled-cluster methods. These techniques are extraordinarily useful, but they contain a basic trade-off. The cost of the most accurate methods rises so rapidly with the number of electrons that they cannot be run on large molecules at all. Chemists therefore use cheaper approximate methods that retain an error of a few kcal/mol. Because a difference of one or two kcal/mol in binding energy can determine whether a drug works, that error is commercially expensive.

The quantum possibility. A quantum computer represents quantum systems naturally. Interactions among electrons in a molecule are a quantum problem. In principle, therefore, a sufficiently large error-corrected quantum computer can perform electronic-structure calculations that are classically impossible. This is the strongest theoretical case for quantum computing, and it deserves full credit in this article.

But exaggeration must be avoided. Drug discovery is not simply a matter of solving molecular equations. A drug must be absorbed by the body, reach the correct location, survive metabolism, avoid toxicity, remain stable, be manufacturable at scale, and prove safe and effective in clinical trials involving thousands of patients. Most of these stages concern biology, regulation and human variability, not computation. Even if a quantum computer improves molecular simulation tenfold, it will not compress the entire drug-development timeline as dramatically as headlines suggest. It will remove one bottleneck, not the whole pipeline.

Where the economic value lies. In two places. First, eliminating failed candidates earlier, because late-stage failure is the most expensive. Second, understanding systems such as catalysts and enzymes that involve transition metals, where classical methods are weakest.

Business model. Today, this consists primarily of research partnerships, not product sales. Boehringer Ingelheim partnered with Google Quantum AI in 2021, one of the earliest formal pharma-quantum partnerships, and later expanded to IBM Quantum. AstraZeneca published peer-reviewed results with Quantinuum on dioxygen binding. Sanofi worked with SandboxAQ’s biological-simulation platform. Cleveland Clinic launched a ten-year Discovery Accelerator with IBM.

Today’s maturity: Pilot and Research stage. Commercially relevant molecules require hundreds to thousands of error-corrected logical qubits. The highest verified figure today is 96. The gap spans several orders of magnitude.

The biggest obstacle. It is not only the number of qubits. Another major obstacle is that the answer produced by the quantum computer cannot be verified classically, because if such verification were possible, the quantum computer would not be needed. The pharmaceutical industry is highly regulated, and every decision requires evidence. Winning acceptance for a calculation that cannot be independently checked is, by itself, a major challenge.

2. Materials Science

The problem. Many of the modern economy’s greatest bottlenecks are materials bottlenecks. Batteries are not good enough, limiting the range of electric vehicles. There is no room-temperature superconductor, so electrical transmission loses energy. There is no inexpensive carbon-capture material, so reducing emissions remains costly.

Today’s solution. Experimental discovery, which is slow and expensive, and classical simulation, which is limited for complex materials. The limitations are especially severe for strongly correlated materials, in which electrons interact so powerfully that ordinary approximations break down. These are precisely the materials that display the most interesting behaviour, including high-temperature superconductivity.

The quantum possibility. This may be the most credible long-term application of quantum computing. Many experts believe the first genuine commercial advantage will appear here, even before pharmaceuticals, because materials work often does not require perfect accuracy; knowing which direction is promising may be enough.

Why the economic value is so large. A single breakthrough can be worth more than the entire industry. If the energy density of a battery improves by 30 per cent, the economic effect is measured in trillions of dollars. That is why chemical and energy companies invest in quantum even when it gives them no result today. It is not a lottery ticket; it is insurance.

Business model. Research contracts, joint-development agreements and, in the future, the most attractive form: sharing intellectual property. If a quantum company contributes to the discovery of a catalyst, it may ask for a stake in the patent.

Maturity: Research to Pilot. Google’s Quantum Echoes work points in this direction because it was presented as a tool for understanding the molecular structure of materials such as polymers and battery components.

3. The Chemical Industry

The chemical industry has one feature that makes it particularly interesting for quantum: a very small percentage improvement can have enormous value because the scale is vast.

The best example is nitrogen fixation. The Haber-Bosch process, which makes ammonia from atmospheric nitrogen and supports the world’s food supply, consumes one to two per cent of all global energy because it requires high pressure and high temperature. Some soil bacteria, by contrast, perform the same task at ordinary temperature and pressure using an enzyme called nitrogenase. We still do not fully understand how the enzyme works because its active centre contains an iron-molybdenum cluster whose electronic structure lies beyond the reach of classical methods.

This is the “moonshot” example of quantum chemistry. If a quantum computer can explain the mechanism of nitrogenase and thereby yield an inexpensive industrial catalyst, its value would be measured at the scale of global energy consumption.

Honesty is essential here. The example has been repeated so often in the industry that it has almost become a slogan. Resource estimates suggest that even this one calculation would require millions of physical qubits and a very long runtime. It is a valid goal, not an immediate business plan.

Nearer-term opportunities: small improvements to existing industrial processes, optimisation of separation processes and a better understanding of reaction pathways. BASF, Dow, Mitsubishi Chemical and JSR are among the companies that have launched quantum programmes, primarily to learn and position themselves.

4. Energy

Quantum discussions in the energy sector contain more marketing contamination than most, so the claims must be separated carefully.

Where the argument is strong. Battery chemistry, new materials for solar cells, fuel-cell catalysts and sorbent materials for carbon capture. All of these fall under materials science, and the entire argument above applies to them.

Where the argument is weak. Grid optimisation. Operating an electricity grid is a huge optimisation problem; that is true. But it is primarily a problem in linear and mixed-integer programming, fields in which classical solvers have developed for fifty years and become extraordinarily capable. The real bottlenecks for a grid operator are data quality, communication delays and regulatory constraints, not computation.

Where the argument remains speculative. Fusion. Quantum simulation may play a role in understanding plasma behaviour, but the main obstacles to fusion today are materials engineering, magnet technology and plasma control, not calculation. It is far too early to credit quantum computing for the success of any fusion company.

Oil and gas. Claims are made about quantum applications in seismic imaging and reservoir simulation, but these are principally classical-HPC problems. Interestingly, quantum sensing has a more realistic role in this industry than quantum computing. We will return to that later.

5. Finance

Banks are among the most vocal experimenters with quantum computing. It is important to understand why, because their reason is not what it seems.

Potential applications. Portfolio optimisation, in which the best combination must be chosen from thousands of assets under constraints. Derivative pricing, which uses Monte Carlo simulation and for which quantum amplitude estimation theoretically offers a quadratic speedup. Risk calculations, especially Value at Risk and stress testing. Credit scoring and fraud detection, although these are mainly machine-learning problems and have only a weak relationship with quantum.

The problem with quadratic speedup. Quantum amplitude estimation can reduce Monte Carlo error to the same level in fewer steps. That is mathematically correct. But a quantum gate is thousands or millions of times slower and more expensive than a classical operation, and error-correction overhead comes on top. The point at which quantum becomes genuinely cheaper therefore arrives only for extremely large problems. In many published analyses, that crossover is so distant that today’s commercial derivative-pricing problems come nowhere near it.

So why do banks invest? There are five honest reasons.

First, option value. A large bank has an IT budget measured in billions of dollars. Spending ten or twenty million on quantum is risk management. If the technology succeeds and the bank is unprepared, the loss could be far greater.

Second, talent. A bank with a quantum team can attract PhD-level mathematicians who might otherwise go elsewhere. Those people do much more than quantum work.

Third, learning. An organisation that starts from zero when the technology is ready will be left behind.

Fourth, quantum-inspired classical benefits. This is the least discussed and most real gain. In attempting to reformulate a problem through a quantum lens, teams often discover a better classical solution. That benefit is available today.

Fifth, and it should be acknowledged, publicity. A quantum press release makes a bank look modern. That is not inherently wrong, but it is wrong to mistake it for operational R&D.

Now for the question asked most often. Will a quantum computer be able to predict the stock market?

No.

And the reason is not a lack of computing power. Markets are not unpredictable because the calculations are difficult. They are unpredictable because prices already incorporate the information available to participants, and because markets are self-referential. If someone discovers a pattern that predicts the future, trading on that pattern causes it to disappear. This is a problem of information and reflexivity, not computing power. A faster optimiser cannot manufacture an advantage where none exists.

A quantum computer may balance a portfolio better under constraints. It cannot tell you which share will rise tomorrow.

6. Logistics and Supply Chains

The problem. A delivery company may have thousands of packages, hundreds of vehicles, delivery windows, limits on driver working hours, vehicle-capacity constraints and traffic. The number of possible route combinations grows so rapidly that testing all of them is impossible.

The truth about complexity. The vehicle-routing problem is indeed NP-hard. But there is a crucial misunderstanding that must be cleared up.

The fact that a problem is computationally difficult does not mean a quantum computer will solve it better.

This is the industry’s most common logical error. The reasoning goes like this: the problem is difficult for classical computers; quantum computers are powerful; therefore, quantum computers will solve it. The conclusion sounds reasonable and is wrong.

Quantum computers are not known to solve NP-complete problems efficiently. Grover’s algorithm offers only a quadratic speedup for general search, and much of even that is consumed by error-correction overhead. Shor’s exponential speedup exploits the special mathematical structure of factoring, and routing does not have that structure.

The truth on the other side. In practice, routing is already solved very well. Modern classical heuristics produce, in a few seconds, a solution within one or two per cent of the theoretical optimum. The final two per cent is not the delivery company’s real bottleneck. The real bottlenecks are wrong addresses, incomplete traffic forecasts, customers who are not at home, drivers who fall ill and packages that arrive late at the warehouse.

So what can a quantum company actually sell? This question is the practical centre of the article, and it is answered in detail in a separate section below.

7. Manufacturing

Factory scheduling, deciding the sequence of robotic cells, quality inspection and product-design optimisation all belong to the optimisation category, and the entire argument about logistics applies to them.

One field where the case is relatively stronger is material selection within design optimisation, because quantum simulation enters the picture, not merely combinatorial optimisation. The interest of companies such as Airbus and Boeing is concentrated here, on lighter and stronger alloys and composites.

Another field that receives less attention but is more concrete is the use of quantum sensing in manufacturing. This includes measuring minute magnetic or electric fields in semiconductor fabrication, non-destructive testing for cracks in materials, and measuring currents inside battery cells from the outside. It is happening today, and it is not quantum computing.

8. Automotive

The automotive industry’s quantum interest is real in three areas and exaggerated in two.

Real: battery chemistry, lightweight materials and catalysts. These belong to materials science. Volkswagen, Daimler, BMW, Hyundai and Toyota all have quantum-chemistry programmes of one form or another.

Exaggerated: traffic optimisation and autonomous driving.

A well-known traffic-optimisation demonstration used a quantum annealer to calculate bus routes in a city. It was technically impressive. But classical solvers produced the same or better results, and the project never became a continuing commercial operation.

The bottleneck in autonomous driving is not computational speed. It is perception, safety validation, behaviour in rare situations and regulatory approval. Quantum computing has no established role in any of these. When a company speaks of “quantum AI for self-driving,” it is reasonable to suspect that it has attached a quantum label to an ordinary AI product.

9. AI and Machine Learning

Quantum Machine Learning is the most overmarketed part of the entire field, so it should be examined calmly.

The theoretical promise. Some linear-algebra tasks, such as solving very large systems of linear equations, have quantum algorithms that promise exponential speedups. Since machine learning is fundamentally linear algebra, it is natural to imagine that quantum computers will accelerate AI dramatically.

The first wall: the data-loading problem. A quantum algorithm is fast only when the data already exists in a quantum state. Your data, however, is classical and sits on a hard drive. Loading N data points into a quantum state generally requires operations proportional to N. If loading takes as long as completing the entire job classically, where is the speedup? The proposed solution is QRAM, a quantum memory that remains theoretical and for which no practical construction route is clear.

The second wall: dequantisation. In recent years, researchers have repeatedly shown that classical algorithms can be constructed that are almost as fast as quantum algorithms claiming exponential speedups in QML, under comparable assumptions. It was a serious theoretical setback that is scarcely discussed in popular articles.

The third wall: the scale gap. A modern AI training cluster contains thousands of GPUs performing trillions of operations per second on enormous datasets. Today’s best quantum processors run a few thousand gate operations on a few hundred noisy qubits. The two are not even comparable.

Will quantum computers replace GPUs? In practical terms, no. These are two different technological revolutions often spoken of together merely because both sound like “the future of computing.” AI is a revolution of data and scale. Quantum is a revolution in specific mathematical structures. Their economics, customers, bottlenecks and timelines are different.

Where the two genuinely meet is in the opposite direction. AI is helping quantum computing today, not the other way around. Machine learning is used to build error-correction decoders, calibrate qubits, optimise pulse sequences and simulate quantum systems classically. One of the roots of the Flatiron Institute’s classical algorithm that challenged D-Wave’s claim was the AI-derived method of belief propagation.

One long-term possibility is that quantum computers generate chemistry training data that would be impossible to create experimentally, and classical AI models are trained on that data. In that arrangement, the quantum computer is not a competitor to AI but a supplier. It is an interesting commercial model on which very little work has yet been done.

10. Cybersecurity: The Place Where Quantum Is Making Real Money Today

This is the most important section of the entire article because this is the only field in which quantum technology is generating hundreds of crores of rupees in real revenue this year. Ironically, most of that revenue comes from products that run on classical computers.

What Today’s Encryption Rests On

When you open a banking app, your phone and the bank’s server perform two separate tasks.

The first is key exchange. Both sides have to agree on a shared secret key without sending it openly across the channel. This uses public-key cryptography, principally RSA or elliptic-curve cryptography. Their security rests on a mathematical imbalance: multiplying two large prime numbers is easy, but recovering those primes from the product, known as factoring, is extraordinarily difficult for classical computers. Breaking RSA-2048 classically would take longer than the age of the universe.

The second task is encryption. Once the shared key has been established, the actual data is encrypted with a symmetric algorithm such as AES.

Shor’s Algorithm

In 1994, Peter Shor showed that a sufficiently large and sufficiently accurate quantum computer could perform both factoring and discrete logarithms exponentially faster. The direct consequence is that RSA and ECC, which support almost the entire public-key system used today, would break in the presence of such a machine.

This is not a prediction. It is a proven mathematical result. The only question is when such a machine will be built.

How the Numbers Are Changing, and Why That Is the Most Worrying Part

In 2019, Craig Gidney and Martin Ekerå estimated that breaking RSA-2048 would require roughly twenty million noisy qubits and eight hours. That sounded so remote that people relaxed.

In May 2025, Gidney published a new paper. The revised estimate was fewer than one million noisy qubits and less than one week. That is twenty times fewer qubits than in his own previous calculation. The reduction did not come from new hardware. It came from three algorithm-level improvements: approximate residue arithmetic, yoked surface codes and magic-state cultivation.

And the sequence did not stop there. In February 2026, a research team proposed an architecture using qLDPC codes and claimed that the same task could be completed in roughly one month with fewer than one hundred thousand physical qubits. In March 2026, another group proposed using still fewer qubits in exchange for a much longer runtime.

Honesty matters here. All of these are theoretical resource estimates, not built machines. Each estimate depends on assumptions about gate-error rates, cycle times and decoder speed. No existing machine comes close.

But the trend is clear, and its economic meaning is serious: the threat timeline is moving closer even faster than the hardware is improving, because the algorithms are improving too. For any CISO, that is the most uncomfortable fact.

“Harvest Now, Decrypt Later”

This is the logic that makes the market real today.

An adversary does not need a machine capable of breaking RSA today. It needs only to capture encrypted data today and store it. When the machine arrives ten years later, the data can be opened.

That means the confidentiality lifetime of your data is the real timeline. If a medical record, diplomatic message, weapon-system design or long-term trade secret must remain confidential until 2040, its security must be decided today. That is why defence and intelligence agencies were the first to act.

Post-Quantum Cryptography, and the Point Most Often Misunderstood

Post-Quantum Cryptography does not require a quantum computer.

This sentence corrects the most common misconception in the entire subject. PQC is a family of new mathematical algorithms that runs on ordinary laptops, servers and phones. Its security does not rest on factoring or other problems that Shor can break, but on lattice-, code- and hash-based problems for which no efficient quantum algorithm is known.

Quantum is in the name because the threat is quantum. The technology itself is entirely classical.

The state of standardisation. On 13 August 2024, the US National Institute of Standards and Technology published three final standards: FIPS 203 for ML-KEM key establishment, FIPS 204 for ML-DSA digital signatures, and FIPS 205 for hash-based SLH-DSA signatures. On 11 March 2025, NIST selected HQC as a fifth algorithm, a code-based backup that provides a different mathematical foundation in case a weakness is found in lattice mathematics. Its draft standard was expected in 2026 and finalisation around 2027. FN-DSA, based on Falcon, was being developed as FIPS 206.

Timelines that are now becoming legal obligations. Under NIST guidance, quantum-vulnerable algorithms at the 112-bit security level will be deprecated after 2030 and disallowed after 2035. In June 2026, an executive order in the United States imposed stricter deadlines on high-value federal systems: the end of 2030 for key establishment and the end of 2031 for digital signatures. Federal contracting rules are also being amended to require contractors to comply with NIST PQC standards. Under the NSA’s CNSA 2.0 policy, new acquisitions for national-security systems should be quantum-resistant from 2027, with a complete transition mandatory by 2035. In June 2025, the European Union published a coordinated roadmap asking member states to establish national PQC strategies and begin cryptographic inventories by the end of 2026, migrate high-risk critical infrastructure by 2030 and complete the transition by 2035.

Now read this in the language of business.

A contractor must comply to win a contract. A bank must answer to its regulator. A hospital must demonstrate the long-term security of patient data. This is no longer “something it would be good to do.” It is becoming a condition of purchase.

And that is how a real market, one that exists today, is created.

PQC Migration Is a Service Industry

For a large organisation, moving to PQC is not a software update. It is a multi-year programme, and it creates at least six distinct commercial products.

First, cryptographic discovery and inventory. Most large organisations do not know where or which cryptography they are running: every TLS certificate, every SSH key, every VPN, every code-signing key, every hardware security module, and every old application into which someone added a library ten years ago. Tools that discover and list these assets, producing what is called a cryptographic bill of materials, are a clear software product.

Second, prioritisation and risk assessment. Not every system needs to change at the same speed. Data that must remain confidential for a long time has to move first. This is consulting work.

Third, crypto-agility architecture. The real long-term product is a system designed so that algorithms can be changed easily in the future. This is an architectural service whose demand will remain even after the PQC transition.

Fourth, product upgrades. HSMs, certificate authorities, VPN equipment and firmware for IoT devices. For hardware and software vendors, this creates a refresh cycle, and refresh cycles are among the most dependable sources of revenue in the technology industry.

Fifth, testing and certification. The keys and signatures used by the new algorithms are significantly larger than the old ones, creating real problems for network packets, embedded devices and bandwidth-constrained systems. Testing them is a specialist service.

Sixth, government and defence contracts. This is the largest and most stable component.

This market exists today, its customer is clear, its budget is approved, and it does not require a single qubit. SandboxAQ, PQShield, ISARA, Entrust, Thales, IBM and Palo Alto Networks, as well as almost every major consulting firm, are participating in it.

If you take only one commercial conclusion from this entire article, it may be this: the clearest money being made from quantum technology today does not come from building quantum computers, but from selling protection against the future threat they create.

Quantum Key Distribution, and the Question That Must Be Asked

QKD is a completely separate technology and must be distinguished from PQC.

In QKD, two parties create a shared key by sending single photons across an optical-fibre or satellite link. Under the laws of quantum mechanics, anyone attempting to measure the photons in transit must alter their states and reveal the interception. QKD’s security therefore rests not on mathematical difficulty, but on the laws of physics. In theory, it is extremely attractive.

So why will the entire internet not move to QKD tomorrow?

The question captures the spirit of this article perfectly, and the answer lies in five practical obstacles.

First, it solves only key distribution. QKD does not encrypt, sign or authenticate. Most importantly, QKD itself requires classical authentication to ensure that the party at the other end is truly the intended party. Without it, a man-in-the-middle attack remains possible. In other words, QKD does not eliminate classical cryptography; it depends on it.

Second, distance. Photons attenuate in fibre. On ordinary telecom fibre, the practical limit without amplification is a few hundred kilometres. Beyond that, “trusted nodes” are required, points at which a key is decrypted and retransmitted. Every trusted node is another point that must be trusted, weakening the overall security guarantee. The genuine solution is a quantum repeater, which does not yet exist commercially.

Third, infrastructure and cost. QKD needs dedicated fibre or a special wavelength channel, together with specialised hardware at both ends. Reports have placed the cost of an enterprise QKD node at around $100,000, and the hardware alone for a five-node metropolitan network above $500,000, before installation and maintenance. PQC, by comparison, is a software-library update with an almost zero marginal cost.

Fourth, denial of service. If an attacker floods the fibre with bright light and saturates the detectors, key distribution stops. Security remains intact, but service does not.

Fifth, implementation flaws. QKD security is proven for ideal equipment. Real equipment has side channels, and the research literature contains several examples of successful attacks on commercial QKD systems.

Governments are divided, and the division is revealing. The US National Security Agency has stated clearly that it does not support QKD for national-security systems and does not expect to certify such products until these limitations are addressed. Britain’s National Cyber Security Centre has also declined to support QKD for government or military use and identifies PQC as the primary defence. Germany’s BSI and France’s ANSSI have likewise urged caution in a joint position paper.

The European Union, on the other hand, is building continent-wide quantum-communication infrastructure through EuroQCI. China has invested heavily in the Beijing-Shanghai backbone and the Micius satellite. In India, under the National Quantum Mission, indigenous technology from QNu Labs was used in April 2026 to demonstrate a one-thousand-kilometre quantum-secure communication network, ahead of an eight-year target of two thousand kilometres.

Why the contradiction? Because these countries are pursuing more than cost efficiency. Their goals include technological sovereignty, long-term strategic capability and the development of domestic industry. A government may buy something that would be economically irrational for a company.

11. Quantum Communication and the Quantum Internet

Beyond QKD lies a larger vision called the quantum internet. Before explaining it, we must be clear about what it is not.

It is not a faster replacement for today’s internet. You will not watch videos on it. Entanglement does not make data travel faster.

What it is. A mechanism for distributing entanglement between two distant points. It requires sources of entanglement, quantum memory capable of preserving a quantum state for some time, and quantum repeaters that extend distance through entanglement swapping without requiring a trusted node in the middle. Quantum repeaters are the greatest unresolved engineering challenge in this vision. No commercial quantum repeater exists today.

If it is built, what will it be used for? There are four genuine uses.

First, device-independent cryptography, which could remove the implementation-related weaknesses of present QKD.

Second, distributed quantum computing. If millions of qubits cannot be placed on one chip, several smaller machines could be joined through quantum links to form a larger one. This may be the most important long-term commercial reason because it addresses the fundamental scaling problem. IonQ’s acquisition of Lightsynq and IBM’s plan for module-connecting L-couplers both point in this direction.

Third, blind quantum computing, in which a customer could run a calculation on a quantum computer without revealing what the calculation is. For a pharmaceutical company, this would be extraordinarily valuable because its molecule may be its most valuable asset.

Fourth, distributed quantum sensing, in which distant sensors are entangled to increase their combined sensitivity. It has possible applications in telescope arrays for astronomy and in geological monitoring.

Who will the first customers be? Governments, defence agencies, national laboratories, central banks and financial infrastructure, as well as telecom operators seeking to increase the future value of their fibre assets. This is a B2G and B2B market, not a consumer market, and it may never become one.

12. Quantum Sensing: The Part That Does Not Need a Quantum Computer at All

If you are looking across this entire industry for the most credible source of real, repeatable, product-based revenue in the near future, the strongest answer is not quantum computing but quantum sensing.

The reason is beautifully simple. In quantum computing, we must fight decoherence because we want to isolate the qubit completely from the outside world. In quantum sensing, we use decoherence. We want the qubit to be affected by the outside world because we learn about that world by measuring the effect.

That is why the engineering challenge in sensing is several orders of magnitude easier than in computing. You do not need millions of qubits and error correction. You need a small number of well-controlled atoms or spins.

The Principal Technologies and Their Actual Maturity

Atomic clocks. The most mature category. They use the hyperfine transitions of rubidium or caesium atoms. Chip-scale atomic clocks are now mass-market commercial products and are regarded as the most widely deployed quantum sensors in defence. Their Technology Readiness Level is about 7 to 8, meaning real deployment in the field.

Magnetometers. There are two principal types: optically pumped magnetometers, which measure spin precession in alkali-vapour cells, and nitrogen-vacancy diamond sensors, which use a particular defect in diamond, operate at room temperature and can be reduced to chip scale. Their Technology Readiness Level is about 6 to 7, meaning commercial prototypes in limited field use.

Gravimeters. Absolute measurements of gravity through cold-atom interferometry, linked to fundamental constants rather than to a mechanical spring that must be recalibrated repeatedly. Their Technology Readiness Level is about 5 to 6, meaning pre-commercial field trials.

Inertial sensors. Atom-interferometry-based accelerometers and gyroscopes designed to determine position without GPS.

Rydberg RF sensors. Highly excited atoms used to measure electromagnetic signals, without antenna size being tied to wavelength.

Real Commercial Uses

Underground mapping. A gravimeter detects differences in density beneath the ground. That means it can reveal an old mine, cave, tunnel, water reservoir or ancient foundation before anyone digs. Unexpected underground obstacles are a major source of delay and cost overrun in construction. This creates a direct, measurable ROI.

Mineral exploration. For mining companies, knowing where to dig is a billion-dollar question. That is why the investment arm of mining giant BHP has invested in Atomionics, a maker of cold-atom gravimeters. Companies such as Nomad Atomics also work in this field, while established companies such as GEM Systems have sold thousands of optically pumped magnetometers for geophysics and archaeology over several decades.

Infrastructure inspection. Non-destructive inspection of bridges, pipelines and underground utilities.

Medical sensing. This is the most promising civilian market. Wearable magnetoencephalography systems based on optically pumped magnetometers measure the brain’s electrical activity without the liquid helium required by SQUID-based systems, in which the patient must remain still. The new systems move with the patient’s head, making examination practical for children and people with epilepsy. Work is also under way with Mayo Clinic on imaging magnetic signals from the heart, while NV-diamond devices are being explored for detecting rare cells in blood.

Defence. Detecting submarines, locating hidden tunnels and unexploded ordnance, and navigating through magnetic anomalies.

Why Its Economics Are Better Than Quantum Computing

The product sold is a physical instrument. A customer buys a box, installs it on a truck, drone, aircraft or in a hospital, and it works. There is no cloud queue, no error correction and no promise that “it will become useful in the future.”

The value proposition is measurable. “Our gravimeter will show you what lies beneath this site before you excavate” is a statement on which a purchasing department can make a decision.

It creates a path to Data as a Service. This is the most interesting commercial innovation in the field. Instead of selling the sensor, many companies sell the measurement. A mining company does not want a gravimeter; it wants to know what is underground. The sensing company therefore operates the equipment, performs the survey and sells the data or analysis. The model is better for three reasons: the customer avoids capital expenditure, the company maximises the utilisation of its expensive equipment, and technical expertise stays inside the company, where it remains a competitive advantage.

An honest note about market size. Estimates from different market-research firms vary widely and should be read with caution. One estimate placed the quantum-sensor market above roughly $410 million in 2025 with double-digit annual growth possible over the following decade, while other analyses projected as much as $2 billion by 2030. These figures depend heavily on what is included in the definition of a “quantum sensor.” They may look small beside projected trillions for quantum computing, but there is an important difference: the sensing figures are based on actual sales, not estimated future economic value.

13. GPS-Independent Navigation

This is the fastest-growing submarket in quantum sensing, and the reason is geopolitical rather than technological.

The problem. Almost every moving part of modern civilisation depends on GPS. Yet a GPS signal is extremely weak because it arrives from twenty thousand kilometres away. Jamming it is cheap, and spoofing it, sending a false signal that tells an aircraft the wrong position, is technically possible. In recent years, incidents of GPS interference in civil aviation have risen dramatically, particularly around conflict zones.

Today’s solution. Conventional inertial navigation systems integrate motion measured by accelerometers and gyroscopes. Their basic problem is drift. Tiny errors accumulate over time, so GPS must repeatedly correct them. Once GPS disappears, uncertainty about position begins to grow.

The quantum possibility. Atom-interferometry-based inertial sensors promise far lower drift because their measurements are tied to fundamental atomic properties, not mechanical components that change over time. A second route is magnetic-anomaly navigation, which uses a pre-existing map of permanent magnetic features in the Earth’s crust and compares it with measurements from a quantum magnetometer.

The market. Military aircraft, submarines, ships, long-range missiles, autonomous systems and, eventually, civil aviation. SandboxAQ’s AQNav and Q-CTRL’s magnetic-navigation product are among the most advanced efforts in this category, and both have received multiyear contracts from the US Department of Defense.

Avoid the exaggeration. These systems are in flight testing and limited deployment today, not mass production. Size, weight and power consumption remain real obstacles, and space on an aircraft is extremely expensive.

Business model. Defence procurement contracts characterised by a long sales cycle, high unit prices, rigorous qualification testing and a relationship that can last decades once the product is approved. These are not the economics of a software startup; they are the economics of an aerospace supplier.

14. Atomic Clocks and Precision Timing: The Quantum Industry Already Managing Trillions

The quietest, and perhaps most important, lesson in this article is hidden here.

The most successful quantum technology is not in the future. It has been operating since the 1950s, and no one feels the need to call it quantum.

An atomic clock is a pure quantum device. It treats the frequency of the transition between two energy levels in a caesium or rubidium atom as the definition of time. The international definition of the second itself is based on this transition.

Now consider where these clocks sit.

GPS. Every navigation satellite contains atomic clocks. GPS is fundamentally a time-broadcast system. Your phone calculates distance by measuring differences in the arrival time of signals. An error of one microsecond becomes a positional error of hundreds of metres.

Telecom. Base stations in a mobile network must remain synchronised; otherwise, handovers break and spectrum efficiency falls.

Financial markets. Regulatory frameworks require timestamps and traceability down to the microsecond for high-frequency trading so that the sequence of trades can be reconstructed later.

The electricity grid. Phasor measurement units compare the phase of electrical waves across the grid. The comparison is meaningful only if all their clocks agree.

Scientific infrastructure. Radio-telescope arrays, particle accelerators and seismology.

Three lessons follow, and all three matter commercially.

First, part of the quantum economy is not in the future but in the present, quietly supporting modern civilisation.

Second, successful quantum products do not advertise that they are quantum. A telecom operator does not buy “quantum-powered network synchronisation.” It buys a timing module with a specified stability. When a technology truly matures, its physics disappears into the specification sheet. That is, in itself, a definition of maturity.

Third, and this is a direct signal for the sensing business: the first large market for next-generation optical clocks, which are far more accurate than current atomic clocks, will be the markets where clocks are already sold. A market in which buyers already exist, budgets are already allocated and only the specification must improve is far easier to enter than a market that must be created from nothing.

15. Medical and Biological Applications

This is the field that requires the greatest caution, because it is the easiest place to misuse words.

What already exists and is based on quantum physics: MRI, which operates through nuclear magnetic resonance; PET scans, which use positron annihilation; laser surgery; and electron microscopy. None could exist without quantum mechanics.

But counting them as part of today’s “quantum industry” is misleading. An MRI manufacturer is a medical-device company. Calling it a quantum-computing company is as wrong as calling a maker of laser pointers a quantum company.

What is real in the new generation: OPM-based magnetoencephalography discussed above, NV-diamond-based biomagnetic sensing, and quantum-enhanced imaging at extremely low light levels, which may be valuable for sensitive biological samples because too much light damages them.

What remains speculation: “quantum computing will cure cancer.” The sentence sounds impressive on a conference stage and has no concrete technical path today. It hides two major claims that would have to be proved separately: that a quantum computer can calculate the relevant biomolecules, and that the calculation changes clinical outcomes. The first is distant. The second has not been established at all.

Part Three: Eighteen Ways to Make Money

We now come directly to the question with which this article began. How does a quantum company actually generate revenue? Each model below is examined through the same lens: product, customer, value, revenue, cost, scalability and risk.

1. Sale of a complete quantum system. The product is an entire machine installed on the customer’s premises. Customers are mainly governments, national laboratories, supercomputing centres and a small number of large enterprises. The value lies in sovereign control, data privacy and research capability. Revenue comes through a large upfront contract followed by annual service and upgrades. Costs are extremely high and installation is highly complex. Scalability is weak because every sale is a project. The risk is that the total number of customers worldwide may be only a few hundred. Finland’s IQM is the clearest follower of this model, with a business resembling the sale of scientific instruments.

2. Quantum Computing as a Service. Shared access through the cloud. Customers are researchers, enterprise innovation teams and developers. The value is that they can experiment without spending millions. Revenue is usage based. The cost is that the machine keeps running whether anyone uses it or not. That is the model’s basic weakness: an expensive instrument with low utilisation.

3. Usage-based compute. The per-task and per-shot prices seen above. It is simple and transparent, but has one problem. Customers find it difficult to estimate the cost because they do not know in advance how many shots will be required. Enterprise purchasing departments hate unpredictable bills.

4. Enterprise subscriptions. An annual platform fee that includes a fixed quantity of access, support and priority. This is better for the vendor because revenue becomes predictable, and better for the customer because the budget is fixed. That is why the industry is moving in this direction.

5. Dedicated capacity. A reserved machine or block of time for one customer. AWS Braket’s reservation rates are a public example, with IonQ Forte priced at $7,000 per hour. The customer is someone who cannot tolerate a queue, usually a government or a large pharmaceutical company.

6. Components and subsystems. Cryostats, lasers, control electronics, detectors, chips and interconnects. This is the industry’s most dependable revenue because it does not depend on which architecture wins.

7. Software licensing. SDKs, compilers, error mitigation and simulation packages. Margins resemble software, but the market is limited because the world’s population of quantum developers is still measured in thousands, not millions.

8. SaaS. An industry-specific tool built on quantum infrastructure, such as a chemistry workflow platform. The customer does not need quantum expertise; the customer needs a result. In the long term, this is the most attractive model because it sits closest to the final user.

9. Consulting. Helping enterprises identify which of their workloads might have quantum potential. This is a real market today, and its most honest form is one in which the consultant often concludes that quantum is not needed yet. The risk is that it is constrained by headcount and therefore does not scale quickly.

10. Integration. Connecting quantum workflows to existing enterprise systems. It sounds dull, which is precisely what makes it valuable, because the real money in enterprise software has always been in integration.

11. Research contracts. A corporation or government pays for work on a specific problem. A large part of quantum-hardware companies’ current revenue comes from here. An important warning: it looks like customer revenue but is economically closer to a grant because it does not naturally repeat or scale.

12. Government and defence procurement. Why this will continue is addressed in a separate section below.

13. Intellectual property and patents. Licensing foundational technologies. Historically, this has been a slow but durable source of revenue in deep tech, and the patent portfolio is one reason behind acquisitions by companies such as IonQ.

14. Quantum-safe cybersecurity services. PQC migration, discovery, crypto-agility and compliance. This is today’s largest and clearest quantum-related market.

15. Quantum-communications infrastructure. QKD equipment, network design, trusted-node management and satellite links. Predominantly government funded.

16. Sensor sales. Physical instruments. Conventional, understandable and working today.

17. Data as a Service. Selling the measurement or insight instead of the sensor. One of the industry’s most practical commercial innovations.

18. Outcome-based contracts. The most interesting model for the future. The customer pays not for quantum compute, but for a solved problem: “We will take twenty per cent of the fuel savings we deliver.” This places all the risk on the vendor and is possible only when the vendor has complete confidence in its advantage. Almost no quantum company can offer this today, and that fact is itself the most honest measure of the industry’s maturity.

Part Four: Unit Economics, the Calculation No One Shows

Popular articles discuss revenue. Almost none discuss cost. So let us do that here.

Where a Quantum-Hardware Company’s Money Goes

Research and development. At these companies, R&D expenditure is several times larger than revenue. That is not unusual; it defines this stage of the industry. But it means the company’s lifeline is its investors, not its customers.

Highly specialised labour. Experienced quantum-hardware engineers number in the hundreds worldwide, not the thousands. Demand far exceeds supply. This cost is rising, not falling.

Fabrication. Manufacturing quantum chips is semiconductor fabrication, with all its capital intensity. IBM moved its principal quantum-chip manufacturing to the 300 mm wafer facility in Albany, and a large part of the May 2026 government agreement was devoted to foundry capacity.

Yield. This one word conceals the largest uncertainty in quantum-hardware economics. Not every manufactured chip works. Not every qubit meets specification. If only a few per cent of the chips on a wafer are usable, unit cost remains extremely high. The semiconductor industry took decades to improve yield.

Cryogenic infrastructure and electricity. The power consumption of a dilution refrigerator remains roughly constant whether it contains dozens or thousands of qubits. The positive implication is that energy cost per qubit falls as qubit count rises.

Calibration and maintenance. This is the least understood operating expense. A quantum computer must be calibrated regularly, sometimes daily. During that time it cannot generate revenue, and specialists are needed to do the work. In utilisation economics, that is a direct deduction.

Helium-3 and supply-chain fragility. The entire superconducting branch of quantum computing depends on a rare isotope whose global supply comes primarily as a by-product of nuclear-weapons programmes. That is why companies are developing plans as extreme as mining the Moon to secure supply.

Sales cycles. An enterprise or government deal can take eighteen to thirty-six months. Throughout that period, the sales team’s costs continue while revenue remains zero.

The Real Numbers, and the Uncomfortable Conclusion They Produce

According to McKinsey’s 2026 Quantum Technology Monitor, the quantum-computing industry as a whole generated more than $1 billion in revenue in 2025 and could reach $4.4 billion by 2028. The same report estimated an economic value of $1.3 trillion to $2.7 trillion by 2035.

Pay attention to the distance between those two figures because it is one of the industry’s most misleading features. “Economic value” and “vendor revenue” are not the same thing. If quantum computing saves a pharmaceutical company $2 billion, that is $2 billion of economic value, but the quantum company may capture only five or ten per cent of it. McKinsey’s own analysis acknowledges that a large share of the value will flow to end-user industries, not quantum vendors.

Now look at the company level. According to reports, IonQ recorded roughly $130 million in GAAP revenue in 2025, more than any other company in the field and the first time a quantum company crossed the $100 million threshold. Quantinuum, by contrast, reported approximately $30.9 million in revenue for 2025, up from $23 million in 2024.

Now place that beside the technical record. Quantinuum has the best logical-qubit encoding ratio in trapped-ion technology and some of the industry’s highest gate fidelities.

In other words, technical leadership and revenue leadership do not belong to the same company. That reveals something deep about the industry: today’s revenue comes not from technical superiority but from government contracts, acquisitions, cloud distribution and arriving early in the market. This is normal in any early deep-tech industry, but investors often assume that the company with the best technology will earn the most money.

Will These Companies Ever Earn Software-Like Margins?

The answer varies by layer, which is the practical conclusion of this entire analysis.

Component suppliers will never earn software margins. They are precision-manufacturing companies, and their economics will remain like the scientific-instrument industry: healthy rather than extraordinary, and dependable.

Hardware companies may never do so, at least not by selling machines. They are too capital intensive. Their true economic path is to turn the hardware into a platform and sell services on top of it, much as printer manufacturers earn money from ink.

Cloud providers will not do so until utilisation is high. Utilisation will not become high until a workload exists that customers want to run repeatedly.

Middleware and software could, potentially. Reproduction costs are zero and switching costs can be created.

Consulting cannot, because it is tied to people, but it provides cash flow that can support everything else.

Sensing occupies an interesting middle position. It is hardware, so margins are limited, but a Data-as-a-Service model can move the company beyond the margin constraints of equipment sales.

Cybersecurity migration is the most attractive. It is software and services, the customer exists today, and regulatory pressure drives demand.

Part Five: Who Will Make Money First

This is not a prediction but an argument. It rests on the principle that the first to earn is the business whose customer exists today and whose value can be measured today.

In the strongest position:

Component suppliers. They are already selling, and their market does not depend on who wins.

PQC-migration companies. Regulatory timelines have become legal pressure, and buyers already have approved budgets.

Atomic-clock and timing companies. Their market is already operating and they need only improve the specification.

Defence-focused sensing companies. GPS interference is a problem today, not tomorrow.

In a middle position:

Quantum-cloud providers, whose revenue is rising but whose fundamental economic question remains unanswered.

Consulting firms, which earn money today but have limited growth.

Civilian sensing companies, which still have to educate the market.

In the most difficult position:

Pure hardware companies, which face the longest road, the highest cost and the greatest technical risk. The winners may become the largest companies in the industry; the losers may disappear completely. It is an asymmetric bet.

Application-software companies waiting for hardware that does not yet exist.

The “Picks and Shovels” Argument, and Where It Breaks

During a gold rush, the person selling shovels always wins. The saying is almost true in quantum. Almost.

Where it works: cryogenics, lasers, control electronics, photonics, detectors, testing equipment, fabrication services, cloud infrastructure, cybersecurity and specialist talent. Demand for these does not depend on which qubit architecture wins.

Where it breaks, and why that matters:

First, the shovels in quantum are architecture specific. A dilution refrigerator is essential for superconducting qubits and almost irrelevant to photonic quantum computing. If photonics wins, a large part of the cryogenics market disappears. In the real gold rush, the same shovel worked in every mine. Not here.

Second, there are very few buyers. Millions dug for gold in California. There are only a few hundred serious quantum buyers worldwide. Losing one customer can ruin a supplier’s entire year.

Third, demand is investment driven, not revenue driven. If quantum funding falls sharply, component demand will fall immediately no matter how good the technology is. Gold miners bought shovels when gold was coming out of the ground. Here, customers are buying while the gold has not yet appeared.

Fourth, the threat of vertical integration. Large hardware companies may begin manufacturing critical components themselves, as has happened repeatedly in the semiconductor industry.

Startup Opportunities: Do You Need to Build Your Own Quantum Computer?

If someone asks, “I want to start a business in the quantum industry. Do I have to build my own quantum computer?”, the answer is clear.

No. In fact, it may be the worst option because it requires the most capital, the rarest talent and the longest wait.

The opportunities can be divided into three levels according to capital requirements.

Low capital, with no need for world-class physicists:

Education and training. The shortage of people who understand quantum is a genuine industry bottleneck.

PQC-migration consulting. This requires knowledge of cybersecurity and enterprise IT, not quantum physics. It is today’s most accessible and clearest opportunity, especially in a country such as India with a large IT-services base, where thousands of organisations need the work and very few people know how to perform it.

Developer tools, documentation, benchmarking and industry analysis.

Technical content and market research, because the field is so complex and so saturated with hype that reliable explanation has a market of its own.

Moderate capital, with some specialised expertise required:

Industry-specific application software and workflow platforms.

Middleware, compilers and error mitigation.

Sensing applications in which the company does not manufacture the sensor, but applies it to a particular industry problem. A company selling underground-survey services to the construction industry does not need to design an atom interferometer; it needs to understand construction.

Integration and systems engineering.

Very high capital, with a world-class scientific team essential:

Quantum processors, fabrication, quantum-networking infrastructure, advanced sensors and cryogenic systems.

At this level, you need hundreds of crores of rupees, a decade of patience and several of the few hundred people in the world capable of doing the work. This resembles a national project more than entrepreneurship, which is why governments are investing directly.

One observation may be the most useful of all: the least competitive opportunities in this industry are those that require less quantum physics and more industry knowledge. Everyone wants to build a qubit. Very few people want to discover what a mining company will actually buy.

Who Writes the Cheque

For every business, this is the final question. The possible buyers are listed below, and each behaves differently.

Governments and defence agencies. The largest and most patient buyers. They purchase before technological maturity because their objective is strategic capability, not ROI. The sales cycle is long and requirements are strict, but the relationship is stable once the company is inside.

National laboratories and supercomputing centres. They buy machines and often connect them to existing HPC infrastructure.

Universities. Smaller budgets but greater numbers, and they train future users.

Pharmaceutical and chemical companies. They participate through research partnerships. Their budget comes from R&D, not IT, and the distinction matters because R&D tolerates more risk.

Banks and financial institutions. They spend from innovation budgets, usually smaller amounts but with faster decisions.

Automotive and aerospace companies. They buy long-term materials research.

Telecom operators. They pay for timing, synchronisation and future quantum networking.

Cloud providers. They are the most interesting buyers because they are both customers of hardware companies and their distribution channel. This double relationship is a strategic risk for hardware companies, just as it is for any vendor dependent on a platform.

Cybersecurity departments. They fund PQC migration. Their motivation is fear and compliance, the two most reliable drivers in a technology market.

Mining, construction and energy companies. They buy sensing and are the most practical customers because they demand a clear ROI and purchase quickly when it is demonstrated.

Part Six: Governments, Geopolitics and Capital

Why Governments Are Investing So Much Money

It is important to understand that the logic of government investment is fundamentally different from the logic of commercial investment. An investor demands a return. A government demands capability.

National security. If an adversary is first to build a cryptographically relevant quantum computer, it may be able to open decades of accumulated encrypted communication. This argument alone is considered sufficient to justify almost any expenditure.

Technological sovereignty. If an economy depends on another country for a critical technology, it is strategically vulnerable. The semiconductor supply-chain experience taught every major economy this lesson.

Military sensing. GPS-independent navigation and submarine detection have direct military value.

Economic competition. The belief that whichever country leads this technology will become the centre of the next industrial wave.

Scientific leadership. Prestige, the ability to attract talent, and the entire ecosystem that forms around a leading research base.

But a serious analytical warning is necessary here. Government investment should not be treated as proof of commercial maturity. Governments routinely invest in technologies that never become commercially successful, and that is part of their role. When a company presents a government grant as evidence of legitimacy, the correct question is how much of its revenue comes from customers who also had the option not to buy.

The Geopolitical Picture

The United States. The Department of Commerce’s $2.013 billion agreement in May 2026 represents a clear change in approach. The government is no longer providing only grants; it is taking equity. It is an extension of the industrial-policy approach adopted in semiconductors. The move also drew criticism, which should be recorded: some reports raised questions about links between a few recipient companies and the administration. DARPA’s Quantum Benchmarking Initiative is a separate and technically important programme whose stated purpose is to determine whether these companies can build an industrially useful quantum computer by 2033. In other words, the government itself is acting as an independent evaluator, a healthy development in a field saturated with hype.

China. Heavy state investment, work on both superconducting and photonic approaches, satellite-based QKD through Micius, and a quantum-communication backbone from Beijing to Shanghai. An important caution is that data from this field is less transparent. McKinsey explicitly notes in its report that information about startup investment in China is limited. Comparative claims must therefore be read carefully.

Europe. The Quantum Flagship programme, the EuroQCI communications infrastructure, and a position on QKD that differs from the United States. Sovereignty is an especially strong element of Europe’s strategy. Germany, France, the Netherlands and Finland have produced strong domestic players, including Pasqal, IQM, Quandela and Alice & Bob.

India. The National Quantum Mission began in April 2023 with an approved budget of ₹6,003.65 crore for the period from 2023-24 to 2030-31. It uses a hub-and-spoke model with four thematic hubs and aims to build systems ranging from 50 to 1,000 physical qubits within eight years, pursuing both superconducting and photonic routes.

Progress has two sides, and honesty requires both.

The positive side: in April 2026, the government announced that indigenous technology from QNu Labs, an IIT Madras-incubated startup funded through the NQM, had been used to demonstrate a one-thousand-kilometre quantum-secure communication network, against an eight-year target of two thousand kilometres. The company’s platform was also independently validated, confirming secure key generation over 200 kilometres of standard telecom fibre without amplification, with a quantum bit error rate below four per cent. The system could coexist on the same fibre with ordinary 10 Gbps data traffic. The number of startups supported under the NQM increased from eight to seventeen across quantum computing, communication, sensing and materials. Andhra Pradesh launched the Amaravati Quantum Valley initiative.

The realistic side: India still imports several critical components needed by quantum laboratories, especially high-quality lasers and cryogenic systems. This is one of the most concrete opportunities for Indian entrepreneurs because it is not a problem of world-leading physics research; it is a problem of high-precision manufacturing, and India has the industrial base to move in that direction. One startup seeking to build precision power supplies for quantum computers received NQM approval within ten days, a positive signal about policy speed.

India’s most practical opportunity may not be building a quantum computer. It may lie in three areas: component manufacturing, global PQC-migration services that play to India’s natural strength in IT services, and the creation of a quantum-skilled workforce.

Standards, Regulation and Export Controls

Cybersecurity standards were examined in detail above. They are the most concrete regulatory force today.

Benchmarking and interoperability. A serious problem is the absence of any universally accepted standard for comparing quantum computers. Qubit count is useless, quantum volume is limited, and algorithmic qubits are vendor specific. The commercial effect is that buyers cannot compare products, and a market in which buyers cannot compare is inefficient. That is why independent benchmarking is itself a commercial opportunity, and why DARPA’s evaluation programme matters.

Export controls. Several countries have placed quantum computers, dilution refrigerators and related components on dual-use technology lists. The direct commercial consequence is that a quantum company’s global market may be legally restricted, and every financial forecast should account for it.

Investment, Valuation and the Bubble Question

Now we come to the question that cannot be avoided.

The numbers. According to McKinsey, investment in quantum reached $12.6 billion in 2025, 6.3 times the previous year’s figure. Roughly 60 per cent was concentrated in the ten largest deals. An especially interesting shift occurred: in 2024, governments and public institutions accounted for about one-third of investment; in 2025, their share fell to only three per cent as private capital and public markets took the lead.

An arithmetic observation. Quantinuum went public in June 2026 at an initial valuation of roughly $15.6 billion, against reported 2025 revenue of about $30.9 million. That is a revenue multiple of roughly five hundred. A mature software company, by comparison, may trade at ten to twenty times revenue.

This alone does not prove that the valuation is wrong. Early-stage deep-tech companies are valued not on present revenue but on the size of the market they may create. It does show, however, that almost the entire price of these shares reflects future expectations, not current performance. It follows that one major delay in the roadmap could seriously reduce the valuation.

So is this a bubble?

The most honest answer is that it is the wrong question. The right question has two parts, and their answers are different.

Is the underlying technology real? Yes. Quantum mechanics is the most successfully tested theory in physics. Error correction has been shown to work. Sensing is selling products today. Shor’s algorithm is a proven mathematical result.

Will every existing company survive, and is every existing valuation justified? Almost certainly not.

Both statements can be true at the same time, and history has demonstrated that repeatedly.

Railways. Nineteenth-century Britain experienced a railway-share mania. Many companies collapsed and investors lost heavily. But the tracks remained, and they transformed the economy.

Fibre optics. Between 1999 and 2001, telecom companies spent billions laying optical fibre. The market crashed, many companies went bankrupt, and their assets were sold for almost nothing. That same “dark fibre” later became the backbone of the internet that carries streaming and cloud services today.

Semiconductors. There were dozens of memory companies in the 1980s. Most disappeared. The industry survived and became enormous.

The most likely quantum scenario is similar: the technology will progress, consolidation will occur, some companies will be acquired by larger groups, some will quietly close, and perhaps three to five hardware players will survive. This period will be both the most painful and the most productive.

What Can Kill a Quantum Startup

Betting on the wrong architecture. If you spend ten years on superconducting qubits and photonics wins, all your intellectual property belongs to a world that never arrived.

Error correction proving too expensive. If the overhead is far greater than expected, a useful machine may become so large that it is economically impossible.

Classical algorithms improving. A single research paper can erase the entire value proposition.

Value arriving too late. This is the most common death. The technology works, but only after the money is gone.

Running out of cash. A quantum-hardware company burns a large sum every month. If capital-market sentiment changes, the next round may never arrive.

Dominance by large companies. IBM, Google, Microsoft and Amazon can tolerate losses in this field for decades. A startup cannot.

Component shortages. Helium-3, specialised lasers and cryogenic connectors. A delay from one supplier can move an entire roadmap.

Talent shortages and departures. The loss of one leading scientist can set a small company back by years.

Export controls and security restrictions. The company’s largest potential market may be closed by law.

And the most common, least discussed cause: failure to find a workload for which a customer will pay repeatedly. A company can succeed technically and die commercially.

Part Seven: Where Real Science Ends and Marketing Begins

Some sentences have been repeated so often in this industry that they now sound true. They need to be corrected one by one.

“A quantum computer will solve every problem.”

No. A quantum computer provides an advantage only on problems for which a specific quantum algorithm exists that can exploit interference. The list of such algorithms is short. For most computations, a quantum computer will be slower than a classical computer because its gates are slow and expensive.

“They will replace ordinary computers.”

No. Your phone, laptop, database server and web server will not become quantum. Quantum offers no advantage for these tasks and many disadvantages.

“They can break every password today.”

No. The largest demonstrations today are at the level of 96 logical qubits. Even the most optimistic theoretical estimates for breaking RSA-2048 require one hundred thousand to one million physical qubits, all of very high quality. The gap spans several orders of magnitude. It is also important to remember that Shor’s algorithm breaks public-key cryptography, not your password. Symmetric encryption such as AES-256 is considered largely secure against quantum attacks because Grover’s algorithm only halves its effective security, a problem addressed by increasing key length.

“They calculate every possible answer at once.”

This is the most widespread and harmful misconception in the subject. Every answer is represented in superposition, but measurement produces only one. The real work is interference, which cancels the wrong answers. If the popular statement were true, quantum computers would solve every NP problem instantly, and we know that they do not.

“Quantum AI will make AI infinitely intelligent.”

No. The data-loading problem, dequantisation results and the gap in scale all argue against this claim. The statement tries to combine two separate technological revolutions and invent a third.

“The more qubits, the better the computer.”

No. Fidelity, connectivity, coherence time and readout speed matter just as much. IBM’s 1,121-qubit Condor did not become its principal product, while 120-qubit Nighthawk did. That is the entire answer.

“Every optimisation problem will become easy.”

No. Quantum computers are not known to solve NP-complete problems efficiently. General search receives only a quadratic advantage, much of which is consumed by overhead.

“A government invested, so it is commercially ready.”

No. Government investment is evidence of strategic importance, not market readiness.

“A large company partnered with it, so it works.”

No. Most enterprise quantum partnerships are exploratory and funded through R&D or innovation budgets, not operating budgets. A partnership means someone is learning, not necessarily buying.

Why Classical Computers Are Not Going Away

The future computing structure will almost certainly look like this:

CPU + GPU + specialised accelerators + QPU

The quantum processor will act as a coprocessor, much as a GPU does today. The main programme will run on a classical computer, which will send a specific subproblem to the QPU. The QPU will return the answer, and the classical computer will continue.

This hybrid model works today and will remain necessary for three reasons. First, most work is better done classically. Second, moving data into and out of a quantum computer is expensive, so it will be called only for a task at which it is genuinely better. Third, error correction is itself a heavy classical calculation, so every quantum computer will remain attached to a powerful classical computer.

The GPU comparison matters commercially as well. GPUs did not replace CPUs. They created a new category, opened a new market, and eventually built a business larger than CPUs, but not by taking their place. That is the most realistic successful future for quantum computing.

Technology Does Not Spread Merely Because It Works

There is a long distance between a technology working and a technology being adopted. Any enterprise adopting a new computational tool needs all of the following:

Clear ROI, in a form the finance department can understand. Reliability, because uncertain results cannot be accepted in a production system. Integration with existing data pipelines and software. Trained staff, who are not currently available. Standards, so the vendor can be changed. Security review, because sending sensitive data to an external cloud is an approval process in itself. Procurement approval, which takes months. Regulatory acceptance in industries such as pharmaceuticals and finance. Predictable pricing, because no department can budget for an uncertain bill.

Apply that list to quantum computing. Today, it satisfies almost none of these conditions. This is not criticism of the technology, but a description of its stage. Every technology passes through this list, and the process takes time.

The Quantum Talent Economy

A common belief is that only physicists have a place in this industry. That is wrong, and in countries such as India the misconception needlessly blocks talented people from entering.

Roles in demand at quantum companies today include physicists and electrical engineers, perhaps the scarcest category because microwave and RF engineering underpins every superconducting system; chip designers; mathematicians; computer scientists, especially for compilers and error correction; software engineers; cryogenic engineers; photonics specialists; mechanical engineers for vacuum systems; product managers able to translate between science and the customer; business strategists; cybersecurity professionals, whose demand is exploding because of PQC; technical writers and educators; and procurement and compliance specialists who understand the world of government contracts.

Most of these roles do not require a deep understanding of quantum mechanics. A good RF engineer can become immediately useful in a quantum company without ever having written the Schrödinger equation.

Part Eight: A Timeline Without Sensationalism

Statements such as “quantum computing will change everything by 2030” are useless because they contain neither specificity nor accountability. It is more useful to divide the future into five categories.

Category One: Commercial Today

Atomic clocks and precision timing. Quantum random-number generators. Magnetometers in geophysics and archaeology. Single-photon detectors. Cryogenic and photonic components. PQC-migration services and tools. Government research contracts. Cloud access for research and education. Limited government QKD deployments. Quantum-inspired classical optimisation software.

These products have customers, they generate invoices, and they make money today.

Category Two: Emerging Now

Field trials of quantum-based navigation. OPM-based medical-imaging instruments. Pre-commercial deployment of cold-atom gravimeters. Error-corrected logical qubits becoming available as a product. Serious hybrid-workflow experiments in large enterprises. Industrialisation of the component supply chain.

Category Three: Possible in the Medium Term, but Uncertain

The first fault-tolerant machines. IBM’s public target is Starling in 2029, the most detailed and openly trackable roadmap in the industry, but it is a target, not a guarantee.

The first narrow but genuine scientific advantage in materials science and quantum chemistry. The first contract under which a company pays for a quantum calculation because the calculation could not be obtained classically.

Category Four: Requiring Major Scientific Breakthroughs

Broad commercial advantage in optimisation, finance and machine learning. Quantum repeaters and a true quantum internet. Million-qubit machines. A cryptographically relevant quantum computer. Commercial blind quantum computing.

Category Five: Still Speculative

Quantum computers replacing general-purpose computing. Consumer quantum devices. “Quantum AGI.” Any claim connecting quantum with general intelligence or consciousness belongs in this category.

Part Nine: Two Examples That Make the Entire Argument Concrete

Example One: A Pharmaceutical Company’s Decision

The situation. Consider a mid-sized pharmaceutical company that we will call Meridian Pharma. It has a candidate molecule that binds to the active site of an enzyme. That active site contains a transition-metal atom, and this is the root of the problem because interactions among electrons in transition-metal systems are so strong that ordinary computational methods become unreliable.

Today’s process. The company runs DFT calculations on its HPC cluster. A complete screening campaign evaluating several hundred variants takes roughly six weeks and, let us assume, costs ₹40 lakh in computing and software licences. That is not the real cost. The real cost is that the results retain an uncertainty of about 3 kcal/mol while the decision needs 1 kcal/mol. The team therefore has to synthesise and test 30 variants in the laboratory, each costing roughly ₹12 lakh and three months. Altogether, one cycle costs about ₹3.5 crore and takes five months.

A hypothetical quantum proposal. A quantum company says its future fault-tolerant system will calculate this active site to an accuracy better than 1 kcal/mol, reducing the number of variants that must be synthesised from 30 to 8.

The savings calculation. Twenty-two fewer variants mean a direct saving of roughly ₹2.64 crore and about two months of time. The real value of that time is far greater because a patent’s term is fixed, and the earlier a drug enters the market, the more revenue it can earn.

Now for the difficult questions.

Does the technology exist today? No. A calculation at this level requires hundreds to thousands of logical qubits. Today’s record is 96.

Can the answer be verified? This is the most serious obstacle. If the answer could be checked classically, quantum would not be necessary. Trust will therefore have to be built in another way, such as benchmarking against known systems or experimental confirmation. Establishing that trust in a regulated industry will take years.

Will classical methods improve in the meantime? Perhaps. Machine-learning-based interatomic potentials are advancing rapidly and some aim to close precisely this accuracy gap.

Meridian’s decision, which is the decision almost every real company makes. It does not buy quantum compute today. It begins a small research collaboration, assigns a two-person team, and signs a multi-year agreement worth a few crore rupees annually, paid from the R&D budget rather than IT.

And that is the real invoice in today’s quantum industry. The invoice is not for compute. It is for access to the future, learning and keeping the option open. Once that is understood, the industry’s revenue figures fall into the correct perspective.

Example Two: An Explanation a Fifteen-Year-Old Can Understand

A delivery company has to divide 3,000 packages among 40 trucks every morning and decide which truck should visit which address, and in what order.

Why is this so difficult? Because the number of choices grows at a terrifying rate. If only one truck has to visit 20 locations, there are so many different possible orders that all the world’s computers could spend their lifetimes trying to count them. Here, there are 40 trucks and 3,000 addresses.

Now ask: what will a quantum company sell to this delivery company?

It cannot sell, “We sell quantum physics.” No one pays for that.

It cannot even sell, “We have a quantum computer.” The delivery company does not need a computer.

What it can actually sell is one of the following:

“Our planning software will reduce the total distance travelled by your trucks by 6 per cent.” Someone can sign a contract based on that sentence because six per cent less fuel has a definite annual value.

“Planning currently takes two hours. We will reduce it to ten minutes, so if a truck breaks down in the morning, you can recalculate the entire plan.” This is not selling speed but resilience, and resilience is often more valuable.

“We will charge twenty per cent of the savings we create.” This is the outcome-based model.

And now for the point that makes the example honest: none of those sentences requires a quantum computer to be inside the product. In most cases today, what actually works is a very good classical algorithm, sometimes inspired by quantum thinking.

The customer does not care. The customer is buying fuel savings, not physics.

That is the most practical lesson in the entire industry. A company that sells its product as “quantum” is selling science. A company that sells it as “6 per cent less fuel” is doing business.

Part Ten: Three Questions That Sort Everything Out

Every time you hear a quantum claim, break it into three separate questions. They are different, and the answer to one does not answer the others.

First: Is it scientifically possible?

Do physics and mathematics permit it?

Second: Is it technically practical?

Can it be built with the instruments, materials and engineering capabilities we have, or are likely to have soon?

Third: Is it commercially profitable?

Will anyone pay enough for it to cover the cost of building and operating it, and will that payment recur?

Now apply the framework to a few examples.

Breaking RSA with Shor’s algorithm: the answer to the first is yes; it is proven. The second is not yet; the machine does not exist. The third question is almost strange because the buyer is a nation-state, not a market.

PQC migration: the first answer is yes. The second is yes; deployment is happening today. The third is yes, and regulators are making it mandatory. Three yeses explain why this is today’s clearest business.

A complete simulation of nitrogenase: the first is yes. The second is not yet; it requires millions of qubits. The third is potentially enormous if the first two are solved.

Mineral exploration with a cold-atom gravimeter: the first is yes. The second is almost; the equipment is still heavy and delicate. The third is yes; mining companies are willing to pay. That is why investment in this field is growing.

Using a quantum computer to predict the stock market: it fails at the first question because the problem is one of information and reflexivity, not computation.

Securing the entire internet with QKD: the first is only partly yes because QKD solves only key distribution. The second is no without repeaters. The third is no because PQC performs the job much more cheaply.

A technology can pass the first question and fail the second and third, and that is not unusual. The history of science is full of technologies that were possible, were even built, but never became businesses.

Conclusion: So What Is the Real Business Model of Quantum Physics?

We now return to the question with which this article began. The answer is not the one most people expect.

Quantum physics does not have one business model because quantum is not one industry.

It is a physical theory from which at least a dozen different industries emerge. Their customers differ, their products differ, their cost structures differ, their maturity differs and their timelines differ.

If this entire article were compressed into one picture, it would look something like this.

At the bottom is science, which receives money from governments and universities and returns knowledge and people. It does not earn profit, and that is not its job.

Above it are components, which make money today, quietly and without headlines, because every experiment needs a cryostat, laser and controller.

Above them is hardware, where the most capital, the greatest talent, the loudest publicity and the highest risk are concentrated, and where no one is yet profitable.

Above hardware are control and middleware, which may quietly become a genuine software business.

Above that is the cloud, which sells access today and may sell calculation tomorrow, if the calculation proves useful.

Above the cloud is software, waiting for hardware that does not yet exist and surviving in the meantime on its classical by-products.

Above software are industry solutions, where the real economic value must be created and where the fewest concrete products exist today.

And running parallel to this entire stack, separate from it, are three businesses that need no quantum computer and are making the clearest money today: cybersecurity migration, sensing and precision timing.

If only one sentence from the entire article is remembered, let it be this: the most mature part of the quantum economy is the part no one calls quantum, and the least mature part is the part everyone calls quantum.

And Finally, the Same Simple Question

Whether you are building a cryostat, writing a compiler, selling a gravimeter, moving a bank to PQC, or dreaming of a million-qubit machine, you must answer the same question that every business has always had to answer:

Which customer problem are we solving for which the customer is willing to pay real money, repeatedly, voluntarily, and at a price higher than our cost?

The question is surprisingly ordinary. A seventeenth-century merchant would have asked it, and the owner of a neighbourhood grocery shop must ask it too.

It is the question that separates a revolutionary scientific achievement from a sustainable business.

Google’s Willow chip is an extraordinary scientific achievement. IBM’s roadmap is a remarkable document of engineering ambition. QuEra’s 96 logical qubits crossed a boundary that many people considered distant only a few years ago. All of these achievements are real and deserve respect.

But none of them is, by itself, an invoice.

That is the most interesting tension in the entire story. Over the last hundred years, quantum mechanics created almost every technology on which the modern world runs, from the transistor to the laser and MRI. All of it happened, and no one called it “quantum business.” It happened because those technologies solved problems for which people were willing to pay, and physics gradually disappeared into the specification sheet.

The second quantum revolution will probably follow the same path. Today, we call it “quantum” because it is new and because the word attracts capital. On the day it truly works, perhaps no one will call it quantum. A mining company will simply buy a better survey. A pharmaceutical company will simply buy a better calculation. A bank will simply buy encryption that will remain secure for the next thirty years.

And then we will know it has become an industry, because people will stop saying its name.