How do quantum computing business models actually work?
In our quantum computing market deck, you will find everything you need to understand the market
SUMMARY
Quantum computing business models work today by selling scarce access to quantum capability through complete systems, cloud capacity, government and R&D contracts, services, and infrastructure software. Dedicated systems and public-sector-backed projects are the strongest models now; recurring cloud usage has the biggest upside if useful quantum computing scales.
The market is much more physical than the usual “quantum cloud” story suggests. Individual machines already sell for millions of dollars, and one hardware order can outweigh months of metered cloud usage.
Governments are not just early customers; in parts of the market they are effectively co-financing the technology roadmap. Rigetti’s extreme government exposure and the A$940 million Australian and Queensland commitment to PsiQuantum show how closely commercial revenue and industrial policy can overlap.
The customer base is widening, but not evenly. IonQ and D-Wave are showing more commercial activity, while smaller suppliers can still depend heavily on laboratories, supercomputing centres and public procurement.
Quantum pricing already spans a huge ladder, from small experiments to six-figure annual capacity commitments and multimillion-dollar systems. That breadth is a sign of an immature market where customers are still buying very different forms of access rather than one standardized compute product.
Recurring revenue exists, but the industry is not yet economically shaped like SaaS. Bookings, backlog, installation milestones and large system deployments can move quarterly results more than steady cloud consumption.
Gross margins can look respectable even while the companies burn cash at extraordinary rates. The problem is less the incremental economics of selling another block of QPU time than the cost of building the next generation of hardware.
Software works best when it solves a hard infrastructure problem across several hardware platforms. Compilers, control systems and quantum error correction have a clearer route to durable value than generic “quantum applications” that still depend on uncertain hardware performance and customer ROI.
AWS and Azure can become powerful distribution layers because they simplify procurement and make multiple QPUs available through one environment. Their leverage grows if hardware performance converges; it weakens if one architecture becomes clearly superior.
The industry is therefore operating in two economic phases. Today, customers mostly pay for scarce infrastructure, research capability and strategic readiness; if fault-tolerant machines deliver clear economic advantage, the centre of gravity should shift toward recurring consumption of useful quantum compute.
This market map, featured in our quantum computing market deck, highlights top companies and startups in the quantum computing market
What do quantum computing companies actually sell today?
Quantum computing companies currently make money from four things: quantum computers, access to those computers, technical work around them, and software that helps customers use the hardware.
The mix varies a lot by company. IQM and Rigetti increasingly look like specialized computer manufacturers because they sell complete systems. IBM sells QPU time through plans ranging from pay-as-you-go access to large annual capacity commitments. IonQ does both: its 2025 filing showed $69.9 million of quantum hardware revenue and $60.1 million from platform access, consulting and support. D-Wave combines cloud access, software, services and full-system sales.
Calling quantum computing a “cloud computing market” gives the wrong picture today. A customer might spend a few hundred dollars running an experiment through AWS Braket, tens or hundreds of thousands on reserved capacity, or several million dollars buying a machine.
The common thread is access to scarce quantum capability. Companies simply package that access in very different ways depending on the customer.
| Business model | What customers pay for | How mature is it today? |
|---|---|---|
| Quantum computer sales | A complete or dedicated quantum system | Already producing multi-million-dollar deals |
| Quantum cloud access | QPU time, tasks, shots or reserved capacity | Real, but still relatively small |
| Development and services | Algorithms, integration, research and training | Important source of revenue today |
| Quantum software | Compilation, control, error correction and development tools | Early, with a few promising infrastructure niches |
Who is actually paying for quantum computing right now?
Quantum computing customers are still concentrated among governments, research centres, supercomputing facilities and large companies willing to spend money before the technology becomes mainstream.
Government exposure can be extreme. Rigetti’s 2025 annual filing says government entities generated 90.2% of its revenue. IQM has sold systems to national and public computing infrastructure, including Oak Ridge National Laboratory in the United States. Australia and Queensland jointly committed about A$940 million to support PsiQuantum’s fault-tolerant quantum computer project in Brisbane, mixing technology policy, infrastructure investment and economic development.
But that dependence does not apply to every company. IonQ’s latest quarterly results are a useful counterexample: around 60% of revenue came from commercial customers, while roughly half came from outside the United States. Its business is becoming noticeably broader than the government-heavy model seen at some smaller peers.
D-Wave is also showing more enterprise activity. In the first half of 2026, nearly half of its revenue came from Forbes Global 2000 companies, according to its latest results. The company also reported that production applications generated 37.3% of its quantum-computing-as-a-service revenue, up sharply from the previous year.
The customer base is widening, but quantum computing still sells best to organizations that can justify spending on strategic capability, research infrastructure or long-term technology readiness. Ordinary corporate IT departments are not yet buying quantum compute the way they buy databases, GPUs or cloud storage.
As this chart shows, and as featured in our quantum computing market deck, search interest in quantum computing has grown significantly
Is quantum cloud access really where the money is?
Quantum cloud access has the cleanest long-term economics, but hardware deployments and large contracts are still doing much more commercial work than the cloud narrative suggests.
D-Wave makes the gap especially visible. Its first-half 2026 bookings reached $35.5 million, yet a single $20 million system sale represented more than half of that amount. The company also generates recurring QCaaS revenue, including growing production usage, but one large machine order can still outweigh a long period of cloud consumption.
IonQ tells a similar story at a much larger scale. Its latest quarterly revenue jumped to $80.1 million, up 287% year over year, and management specifically pointed to record quantum computer deployments alongside cloud utilization. The company is clearly getting paid for access, but physical deployments remain a major part of the growth.
Rigetti has offered quantum computing through the cloud for years, yet its filings say technology-development contracts historically generated most of its revenue. Management expects QPU sales and recurring access to become more important over time.
Cloud usage is already a real business. It just has not become the industry’s dominant economic engine yet.
If you want more recent data on this point, please see our latest quantum computing market report.
How expensive is quantum computing today?
Quantum computing is expensive enough today that serious usage looks closer to scarce supercomputing capacity than normal public-cloud consumption.
IBM currently starts pay-as-you-go quantum access at $96 per QPU minute. Its Flex plan begins at $72 per minute with at least 400 minutes purchased, while Premium starts at $48 per minute with a minimum annual allocation of 5,200 minutes. At the minimum Premium commitment, that works out to roughly $250,000 of quantum compute capacity a year.
AWS Braket shows how much prices vary between hardware providers. IonQ Forte costs $0.30 per task plus $0.08 per shot. An error-mitigated experiment requiring 2,500 shots therefore costs about $200.30. A one-hour exclusive reservation of the same machine costs $7,000.
Other hardware is cheaper. AWS currently lists one-hour reservations at $4,100 for Rigetti Cepheus, $4,000 for IQM Emerald, $3,000 for IQM Garnet and $2,500 for QuEra Aquila.
Azure adds another model. Quantinuum’s Standard subscription is listed at $125,000 per month and Premium at $175,000, while Pasqal charges €3,000 per QPU-hour.
There is already a surprisingly wide pricing ladder, from small experiments to six-figure enterprise commitments.
| Example | Current pricing model | Approximate price |
|---|---|---|
| IBM Pay-As-You-Go | QPU runtime | From $96/minute |
| IBM Premium | Pre-purchased enterprise capacity | From $48/minute, minimum 5,200 minutes/year |
| IonQ Forte on AWS | Task + shot pricing | $0.30/task + $0.08/shot |
| IonQ Forte reservation | Exclusive machine access | $7,000/hour |
| Quantinuum Standard on Azure | Monthly subscription | $125,000/month |
| Pasqal on Azure | QPU runtime | €3,000/hour |
This chart, included in our quantum computing market deck, illustrates yearly VC funding for quantum computing startups
Why are customers buying whole quantum computers?
Customers buy complete quantum computers because dedicated access, local control and integration with supercomputers can be worth more than simply renting QPU time online.
IQM is probably the clearest proof that this market has become real. The company says it has sold 26 full-stack quantum computers and delivered 17. It is also investing more than €40 million to double its cleanroom capacity and support production of up to 30 complete systems a year. Companies do not build that kind of manufacturing capacity for a purely experimental cloud service.
Rigetti has also moved beyond small laboratory systems. It has sold two 9-qubit machines for about $5.7 million combined and secured an approximately $8.4 million order from India’s C-DAC for a 108-qubit system.
Large customers often want the quantum machine sitting next to classical HPC infrastructure. Oak Ridge owns an IQM system on site. IQM has also been selected to integrate a machine with Finland’s LUMI AI Factory. Quantinuum recently announced that Helios will be deployed through Oracle Cloud Infrastructure, bringing the QPU into the same environment as conventional data-centre resources.
That model starts to look familiar once we compare quantum computers with supercomputers, semiconductor equipment or specialized scientific instruments. Some customers prefer to own strategic infrastructure instead of sharing it remotely.
Why pay for quantum computing before it beats classical computers?
Customers pay for quantum computing today because learning, access and strategic positioning already have value even when the immediate workload does not beat the best classical alternative.
A pharmaceutical company that expects quantum chemistry to become useful later has a reason to train researchers and test algorithms now. A national laboratory can justify owning a quantum computer as research infrastructure. A government can treat domestic quantum capability as part of industrial or security policy. None of those buyers needs an immediate “10x cheaper than GPUs” result to approve spending.
There is also a timing problem. If useful fault-tolerant computing arrives, companies will still need algorithms, internal expertise, workflows and integration. Starting all of that after the technology becomes commercially important could leave them years behind.
D-Wave provides one of the better tests of whether this activity can move beyond preparation. Its latest first-half results showed $1.3 million of QCaaS revenue from production applications, equivalent to 37.3% of its total QCaaS revenue. A year earlier, production accounted for only 9.8%.
The absolute amount remains small, so we should not inflate what it proves. But it is useful evidence that some customers are already paying to run operational workloads rather than only demonstrations.
If you want more recent data on this point, please see our latest quantum computing market report.
This chart, included in our quantum computing market deck, looks at IonQ’s strategy in quantum computing
Can a quantum software company work without owning the hardware?
A quantum software company can work without building QPUs, but the safer businesses are sitting deep in the infrastructure stack rather than selling vague “quantum applications.”
Classiq is one of the strongest examples. The company develops software that lets users design quantum programs at a higher level and then target different hardware platforms. When it raised $110 million in 2025, it said both revenue and customer count had tripled year over year and named BMW, Citi, Rolls-Royce, Mizuho and Toshiba among its users.
We still do not know Classiq’s absolute revenue, so that growth claim has limits. The more interesting part is its position in the stack. A hardware-agnostic compiler can stay useful if several QPU architectures survive.
Quantum error correction may be even more attractive. Riverlane’s Deltaflow product handles the decoding and control work required to keep logical qubits alive as systems scale. That kind of technology solves a problem nearly every serious fault-tolerant machine will face.
Services fit naturally around these products for now. Most customers still need help choosing problems, translating them into quantum algorithms and integrating them with classical systems. D-Wave openly describes professional services as a way to move customers toward recurring cloud usage.
The danger is a software company with neither hard technical IP nor a clear route to recurring usage. Zapata, once one of the best-known independent quantum software companies, ceased operations in 2024 after running out of money. Adding a software layer to an immature computing market does not automatically create SaaS economics.
Will AWS, Azure and IBM end up owning the quantum customer?
AWS, Azure and IBM have a strong chance of owning a large part of the customer relationship because quantum computing increasingly fits inside the broader cloud and HPC stack.
AWS Braket already gives one account access to hardware from IonQ, IQM, Rigetti, QuEra and AQT. Microsoft Azure Quantum does something similar with Quantinuum, IonQ, Pasqal and Rigetti. For developers, that is much easier than creating a separate commercial relationship with every hardware company.
The cloud layer can also make different quantum architectures feel more interchangeable. A customer may care about whether IonQ or Rigetti produces the better result, but billing, identity, storage and classical compute can remain inside AWS.
That creates a real risk for QPU manufacturers. Rigetti specifically warns in its filings that cloud providers can influence pricing, promote competing hardware and control how customers discover quantum services.
IBM has a different position because it owns both the quantum hardware and the platform. It can sell free access, usage-based plans, annual capacity and dedicated on-premises systems without giving another cloud provider control over every customer interaction.
AWS or Azure do not automatically capture the best profit pool, though. If one quantum architecture develops a large performance lead, customers will seek that hardware regardless of which marketplace carries it. Cloud distribution becomes much more powerful if QPUs converge toward similar performance.
If you want more recent data on this point, please see our latest quantum computing market report.
This chart, included in our quantum computing market deck, illustrates yearly funding for quantum computing startups
Is quantum revenue recurring yet, or still driven by big one-off deals?
Quantum computing revenue is still unusually lumpy, and anyone reading these companies like SaaS businesses will misunderstand what is happening.
D-Wave is the clearest example. First-half bookings jumped 1,120% year over year to $35.5 million, but $20 million came from one system sale. Even the timing inside the half was uneven: $33.4 million of bookings arrived in the first quarter and only $2.1 million in the second.
IQM shows the same issue through backlog. The company reported only €8.9 million of first-half revenue while its order backlog had climbed above €102 million shortly afterward. A system can be sold well before manufacturing, installation, acceptance and accounting recognition are complete.
Rigetti’s quarterly revenue can also move sharply when a development milestone or hardware order is recognized. Its latest quarter produced $5.1 million of revenue, which remains tiny relative to the size of individual system contracts it is now pursuing.
IonQ is moving toward more scale. Its latest quarter produced $80.1 million and its remaining performance obligations have grown rapidly. Even there, management says deployments are a major driver, so we should still expect more variation than we would see in a mature subscription business.
| Company | Recent revenue signal | What makes the business lumpy |
|---|---|---|
| IonQ | $80.1M latest quarterly revenue | Hardware deployments still contribute heavily |
| IQM | €8.9M first-half revenue | More than €102M of backlog sits ahead of recognition |
| Rigetti | $5.1M latest quarterly revenue | Development milestones and system orders |
| D-Wave | $35.5M first-half bookings | One $20M system order represented most of the increase |
Can quantum computing have good margins?
Quantum computing can already produce decent gross margins on some products, even though the companies themselves remain deeply unprofitable.
D-Wave reported a GAAP gross margin close to 60% in the first half of 2026. Rigetti generated roughly $2.2 million of gross profit on $5.1 million of latest-quarter revenue, putting its gross margin in the low-40% range.
Quantinuum shows why the accounting can get messy. Its latest quarter had a negative GAAP gross margin after compensation and amortization effects, while adjusted gross margin was about 62%. That is a huge difference, so comparisons across quantum companies need a close look at what sits inside cost of revenue.
The basic economics still make sense. Once a QPU has been built and has spare capacity, selling another block of runtime can carry an attractive incremental margin. Software and support can also be profitable.
The difficult part is everything required to build the next machine.
This chart, included in our quantum computing market deck, compares the main business model options for quantum computing hardware startups
Can today’s quantum revenue pay for all this R&D?
Today’s quantum revenue comes nowhere close to funding the industry’s R&D race, so outside capital remains part of the business model.
Rigetti’s latest quarter generated $5.1 million of revenue and a $28.1 million operating loss. Its cash and investments stood at more than $540 million, which tells us more about the company’s ability to keep building than current customer revenue does.
Quantinuum’s latest quarter makes the gap even clearer. Revenue reached $8 million while adjusted EBITDA was negative $68 million. The company could absorb that because its IPO raised $1.7 billion and left it with more than $2 billion in cash and short-term investments.
IonQ has reached much larger commercial scale, but the same pattern remains. Its latest $80.1 million quarter came with an adjusted EBITDA loss above $100 million. The company had billions of dollars of liquidity available and has used that balance sheet to expand through acquisitions as well as internal R&D.
IQM reported €8.9 million of first-half revenue against a €60.5 million operating loss, while holding more than €300 million of cash after its public listing.
This gap is too large to treat financing as a temporary detail. Quantum companies currently sell products to customers while equity investors and governments finance a large share of the underlying technology development.
Does quantum advantage need to arrive for the business model to work?
Quantum companies can build real businesses before broad quantum advantage arrives, but their market stays much smaller if most customers are paying for research and preparation.
We already have enough evidence to reject the idea that nobody will spend meaningful money until fault-tolerant machines exist. IQM has sold 26 systems. Rigetti has multimillion-dollar hardware contracts. D-Wave has landed a $20 million system order. IonQ is now generating tens of millions of dollars per quarter.
Those sales support a legitimate industry around research infrastructure, national capability, specialist computing and early applications.
The ceiling is the problem. Governments, laboratories and large R&D departments can support a sizeable deep-tech sector, but that market is tiny compared with mainstream enterprise computing.
Quantum advantage changes the buyer’s question. Today, many customers are deciding how much they should spend preparing for quantum. Once a machine can solve an important chemistry, optimization, simulation or security problem better than the available classical alternatives, customers start deciding how much workload they should move onto quantum hardware.
That is the point where compute usage could become much larger than technology-readiness spending.
If you want more recent data on this point, please see our latest quantum computing market report.
This chart, featured in our quantum computing market deck, illustrates how revenue is divided among customer segments in the quantum computing market
What happens to quantum pricing if fault-tolerant machines arrive?
Fault-tolerant quantum computing should push pricing toward useful computational output rather than today’s mix of shots, circuits and raw QPU minutes.
Current pricing exposes how experimental the market still is. IonQ can charge per shot on AWS. IBM charges for time during which the QPU is locked. Quantinuum uses Hardware Quantum Credits based partly on gates and shots. Pasqal charges by QPU-hour.
Customers ultimately care about solving a problem. They do not particularly care whether the provider required 10,000 or 10 million physical operations internally to produce the result.
Error correction will make that distinction even stronger. A useful logical computation may require large numbers of physical qubits, classical decoding hardware and continuous error correction. Raw physical resources become an internal engineering problem for the provider.
Mature pricing should therefore concentrate around reserved capacity, runtime, throughput or completed workloads. Large users may sign long-term capacity deals, while occasional users continue buying metered access.
Competition will matter enormously. If several architectures offer similar performance, quantum compute could start behaving like other cloud infrastructure and prices should fall. If one company can solve commercially valuable problems that nobody else can run, it will have far more pricing power.
Which quantum business models look strongest right now?
Dedicated systems and government-backed projects are the strongest quantum business models today, while recurring cloud usage has the biggest upside if useful quantum computing scales.
Hardware sales have become much harder to dismiss. IQM has 26 systems sold and an order backlog above €100 million. IonQ says record computer deployments helped drive its latest quarterly revenue surge. Rigetti and D-Wave are also closing individual system contracts worth millions or tens of millions.
Government and R&D contracts remain powerful because they let companies get paid while developing technology they would have needed to fund anyway. The downside is obvious: procurement is slow, revenue is concentrated and political priorities can change.
Cloud access has better long-term economics. One machine can serve many customers, unused capacity can be monetized and successful workloads can repeat. We simply do not see enough usage yet for QCaaS to carry the cost structure of major quantum companies.
Deep infrastructure software also deserves attention. Error correction, compilers and control systems can become valuable across several hardware architectures without requiring the software company to finance its own QPU factory.
Generic application software looks much harder today. The company needs a commercially useful workload, reliable hardware underneath it and enough differentiation to stop a cloud provider, consultant or hardware vendor from offering the same thing.
| Business model | How strong is it today? | Long-term upside | Biggest weakness |
|---|---|---|---|
| Dedicated quantum systems | Strong | High | Lumpy sales and heavy hardware costs |
| Government / R&D contracts | Strong | Medium | Dependence on procurement |
| Quantum cloud access | Growing | Very high | Usage still too small |
| QEC / compiler / control software | Early but promising | Very high | Depends on quantum hardware scaling |
| Integration and consulting | Useful today | Medium | Labour-heavy |
| Generic quantum applications | Weak to early | Potentially high | Hard to prove customer ROI |
If you want more recent data on this point, please see our latest quantum computing market report.
This chart, included in our quantum computing market deck, shows how cloud quantum computing access technology has evolved over time
So how do quantum computing business models actually work?
Quantum computing companies currently make money by selling scarce access to a technology that customers believe could become much more valuable later, and the industry is using several business models at once because no single one is mature enough to support the whole sector.
The cash coming in today is more physical and project-driven than the popular image of quantum computing suggests. Complete systems are being sold for millions of dollars. Governments pay for research and sovereign infrastructure. Enterprises buy dedicated access, development work and experiments. Cloud platforms meter smaller workloads. Software companies try to capture the layers around the QPU.
The freshest financial data make the split especially clear. IonQ is now growing fast enough that commercial customers account for a majority of revenue, but deployments still matter heavily. IQM has sold far more systems than its current recognized revenue suggests. D-Wave can show rising production usage while a single system contract still dominates bookings. Rigetti remains heavily exposed to government spending. Quantinuum has strong adjusted product margins alongside losses that dwarf revenue.
We therefore see the market in two economic phases.
Today, quantum computing works mainly as a scarce-infrastructure business. Customers pay for machines, capacity, research and expertise, while investors and governments finance the huge R&D bill behind them.
If fault-tolerant quantum computing eventually delivers clear economic advantage, the centre of gravity should move toward recurring compute consumption. At that point, selling millions of hours of useful quantum capacity could become far more important than selling dozens of machines.
That transition is the real business-model bet. Companies are using hardware sales, public contracts, services and early cloud revenue to finance their way toward a market where customers pay because quantum computing solves an important problem better, faster or cheaper than anything else available.
OUR METHODOLOGY
This analysis asks how quantum computing business models actually work today. We separated the market into distinct economic dimensions—hardware sales, cloud access, development work, software, customer mix, pricing, bookings and backlog, production usage, margins, and funding—then looked for the most recent evidence showing actual commercial activity in each one.
We prioritized revenue, system orders and deployments, cloud pricing, production workloads, bookings, backlog and gross margins over broad market forecasts or technical announcements. The aim was to see what customers are paying for now, not to treat future quantum advantage as if it were already commercial reality.
We compared several companies because no single supplier represents the market. IonQ, D-Wave, Rigetti, IQM, IBM, Quantinuum and the major cloud platforms monetize different parts of the stack, while Riverlane helps illustrate the emerging infrastructure-software layer.
We also kept current commercial strength separate from long-term potential. Dedicated systems and government-backed projects can be strong revenue models today even if recurring quantum cloud access or quantum-error-correction software eventually becomes the larger opportunity.
The final two-phase view of the market comes from combining those observations: quantum computing today behaves mainly like scarce research and strategic infrastructure, while a future market with clear quantum advantage would likely shift more of the economics toward recurring consumption of useful compute.
Key sources used for this analysis include IBM Quantum on access models and pricing, Amazon Braket pricing, Microsoft Azure Quantum provider pricing, the Amazon Braket Developer Guide, IonQ’s Q2 2026 results, D-Wave’s Q2 2026 results, D-Wave’s quarterly financial results, D-Wave’s SEC-filed Q2 2026 results, Rigetti’s 2025 Form 10-K, Rigetti’s Q2 2026 results, Rigetti’s C-DAC system order, IQM’s Cineca deployment, IQM Spark system information, the Australian Department of Industry’s State of Australian Quantum report, Quantinuum and Oracle on the Helios deployment, and Riverlane’s Deltaflow 2 technical documentation.
In our quantum computing market deck, we identify pain points entrepreneurs should prioritize
Related blog posts
- Quantum Computing: where's the money now?
- What is the true size of the quantum computing market?
- What does the quantum computing startup landscape look like today?
