Marin Ivezic
IonQ reported $43.1 million in revenue for full-year 2024 and D-Wave reported $8.8 million, both figures taken from the companies’ own annual results releases (IonQ investor relations, D-Wave investor relations). Both are publicly listed, both have been building hardware for more than a decade, and both carry valuations that assume the revenue line bends steeply upward within a few years. That arithmetic determines every quantum investment decision, whether the target is a listed company or a university spinout with four people and a leased lab.
There are two ways to get the assessment wrong, and they fail for the same reason. The first is dismissal: the technology is decades away, the giants will win anyway, nobody is buying. The second is enthusiasm: the science is extraordinary, the team is brilliant, the market will be enormous. Neither position requires anyone to check anything. What follows is a set of checks, built around the claims that come up most often in these rooms and the evidence that either supports them or doesn’t.
Where quantum revenue is recognized today
Before evaluating any forecast, it helps to know what the industry actually bills for. There are four distinct revenue lines, and they behave very differently.
Machine access and hardware sales. Companies sell time on quantum processors, usually through cloud platforms, and occasionally sell whole systems to national labs and government customers. Amazon Braket and Microsoft Azure Quantum both act as storefronts for third-party hardware rather than building all their own. A quantum processor’s capability is usually quoted in qubits, the quantum equivalent of bits. Today’s machines run physical qubits, which are noisy and need constant correction. A logical qubit is an error-corrected unit assembled from many physical ones, and the ratio between them is the central engineering problem in the field. Almost every announced qubit count refers to physical qubits, and a claim that skips the distinction should slow you down.
Sensing, timing and metrology. Quantum sensors measure gravity, magnetic fields, rotation and time with precision that classical instruments can’t reach. This line sells into defence, geophysics, civil engineering and navigation. Contracts are lumpy, often government-funded, and the buyers evaluate instruments the way they always have, against a specification.
Components and subsystems. Cryogenic refrigeration, control electronics, lasers, amplifiers, chip fabrication. These companies sell to the rest of the industry rather than to end users, and their revenue is real product revenue with recognisable gross margins.
Post-quantum cryptography. Post-quantum cryptography (PQC) is classical software designed to resist attack by a future quantum computer. It is sold today, at scale, to regulated industries working through migration deadlines. It gets counted in “quantum market” totals and it shouldn’t be confused with the other three lines, because the technology, the buyer and the sales cycle are all different. If a target’s revenue is mostly PQC consulting, you are looking at a security services business. The migration methodology behind that work is documented in NIST’s post-quantum cryptography guidance.
Sorting a company’s revenue into these four buckets, then splitting each bucket into commercial and grant-funded, usually answers more questions than a market forecast does. We put little weight on published total-market projections for this sector. The forecasts disagree with each other by factors of three or more, they bundle the four lines above into a single number, and they are frequently produced by parties with a position in the outcome.
Four claims you will hear in the room
“It is still ten years away”
This is directionally true for one thing and false for everything else. Large-scale fault-tolerant quantum computing, the machine that would break RSA or simulate a catalyst end to end, is not a near-term product, and any founder who tells you otherwise has told you something useful about the founder.
Everything else in the field is on a shorter clock. Quantum random number generators ship today. Quantum key distribution (QKD), which uses the physics of single photons to detect eavesdropping on a key exchange, is deployed on operational links in several countries. Gravity sensors are in field trials. Machine access is billable now. Lockheed Martin bought a D-Wave One in 2011, and a D-Wave Two went into NASA Ames in 2013 under an arrangement involving Google, NASA and the Universities Space Research Association. Those machines run quantum annealing, a special-purpose technique for finding low-energy configurations of a system, useful for some optimisation problems and not a universal computer. Procurement happened, budgets moved, and that was more than a decade ago – annealing didn’t have to solve those customers’ problems for that to count.
The check to run: which of the four revenue lines does this company sit in, and does its timeline claim match that line? A sensing company promising revenue in eighteen months is making an ordinary instrument-business claim. A gate-based hardware company promising commercial advantage in eighteen months is making a claim the physics does not currently support.
There is also a cost to the ten-years-away position, and biotech shows it clearly. Genentech was founded in 1976, when most established academics considered recombinant DNA far too raw to build a company around. It produced synthetic human insulin in 1978 and licensed it to Eli Lilly, which brought the product to market in 1982, and it went public in 1980. The founders who waited for the science to settle arrived after the platform companies, the patents and the talent had all been claimed.
“Only the incumbents can win”
IBM and Google publish the headline results, so it is easy to assume they will absorb the market. The structure of the industry says otherwise.
IonQ was founded in 2015 by Christopher Monroe at the University of Maryland and Jungsang Kim at Duke, around trapped-ion qubits, in which the qubits are individual charged atoms held in place by electromagnetic fields. In 2021 it became the first pure-play quantum computing company to list publicly, meaning quantum computing is its only business rather than one research line inside a conglomerate. It did that six years after leaving a university lab, competing directly with companies a thousand times its size. PsiQuantum raised a $450 million round in 2021 led by BlackRock. Large incumbents are not crowding startups out of this field. They are buying access to them, hosting them, and funding them.
The supply chain is where the asymmetry is clearest. QuantWare, a spinout from TU Delft, builds and sells superconducting quantum processors to other people’s labs rather than competing for the end-user market. Every hardware roadmap published by a large player increases the number of buyers for components like these. Semiconductors settled into the same shape decades ago, with the best-known brands depending on specialist fabricators for the part that actually determines performance.
The check to run: is this company trying to win the whole stack, or one layer of it? Full-stack quantum computing is a capital-intensive race against companies with unlimited balance sheets. A defensible component, a control system, a fabrication process or a software layer with a real customer list is a different risk profile, and it is frequently the better business.
“There is no customer”
There is no mass market, but there are identifiable customers, and they have names.
Volkswagen ran a bus-routing pilot in Lisbon with D-Wave during Web Summit in November 2019, adjusting routes in something close to real time. JPMorgan Chase has run a quantum research group for years, publishing on portfolio optimisation and cryptography. Utilities, mining companies and civil engineering firms are engaging with gravity sensing because they dig, and they currently dig without reliable knowledge of what is underneath.
What these buyers have in common is that they are not buying a solution. They are buying an option on one, and they are paying for it with budget, staff time and access to proprietary data. For a startup that is enormously valuable, and for an investor it is a signal that needs careful reading. A paid pilot with a named line-of-business owner is evidence of demand. A memorandum of understanding with an innovation team is evidence of curiosity.
The check to run: who inside the customer signed, and out of which budget? Innovation budgets fund experiments and disappear in downturns. Operating budgets fund things people need. Ask which one paid, ask what the customer measured, and ask what happens at the end of the pilot. A pilot with no defined path to a production decision is a marketing expense that the startup happens to be collecting revenue for.
“They can license it later”
This one surfaces mostly around university technology, and it appeals because it sounds prudent. Patent the discovery, let it mature, let a large company license it when the market is ready.
In quantum it works less often than in most fields, for a structural reason. Quantum inventions are rarely a single component that drops into an existing product line. They are usually a combination of materials, device physics, control electronics, calibration procedures and software, plus a set of people who know why the thing works on Tuesdays. A corporate licensee evaluating that package asks whether it has been demonstrated outside the originating lab, whether a prototype exists, whether it integrates with anything, and whether the inventors will be available. Early-stage academic work usually answers poorly on all four. So the patent sits, the maintenance fees accumulate, and the graduate students who understood the calibration procedure take jobs elsewhere.
The successful licensing cases in this field almost always had a company behind them first. Quantum Base, a Lancaster University spinout, has pursued a design-and-license model for quantum authentication tags, but the company still raised money and developed the technology to the point where a manufacturer could implement it. Licensing came after the development work, not instead of it.
The financial comparison also favours company formation for high-impact technology, and the canonical example is not a quantum one. Stanford licensed the PageRank algorithm to Google and took equity alongside the licence. It sold those shares in 2005 for roughly $336 million, a return no royalty schedule would have produced.
The check to run, if you are evaluating a university spinout: has the institution taken equity, and is the licence exclusive, field-limited or something in between? A non-exclusive licence on the core technology is a hole in the moat. A technology transfer office (TTO) that has structured the deal to keep the inventors materially engaged has done something more useful for you than one that maximised the upfront fee.
Two ventures, read the way an investor reads them
A sensing company. Researchers at the University of Birmingham published a gravity-gradiometry result in Nature in 2022, detecting a buried tunnel outdoors, in an ordinary environment, with an instrument built around cold atoms. A gravity gradiometer measures tiny differences in gravitational pull between two points and infers what density sits below. The Birmingham work matters commercially because of where it was done. Laboratory demonstrations of quantum sensing are numerous. Demonstrations outside, against ground vibration and temperature swings and traffic, are not.
Reading that as an investor, the diligence sequence is short. What is the measurement time per survey point, and how does it compare with the incumbent method? What does the instrument weigh, and what does it cost to operate for a day? Who currently pays for underground surveys, how much do they pay, and what does a wrong answer cost them? The physics question was answered by the Nature paper. Everything after that is an instrument business, and instrument businesses are evaluated on cycle time, unit cost and the price of the error the customer is currently absorbing.
A component supplier. The QuantWare model inverts the usual startup question. Instead of asking which end customer will buy quantum computing, it asks who is already spending money building quantum computers. That customer list is knowable, funded, and largely public. Revenue arrives before anyone has demonstrated commercial quantum advantage.
The diligence is correspondingly different. Concentration risk is the first question, because a handful of buyers can account for most of the order book. The second is whether the component is genuinely hard to reproduce or merely currently unavailable. The third is what happens if the field consolidates around one qubit modality, since a component tied to superconducting hardware has a different future than one that serves any architecture. Picks-and-shovels businesses in this sector look safer than they are, and they fail in specific, predictable ways.
The signal hiding in the support question
Some university teams treat quantum commercialization as an ordinary spinout process. Patent, incorporate, join an accelerator, raise. Others go looking for support built specifically for this field, and for an investor that choice carries information.
The specialist programs exist because the general ones were not working. Duality, launched in 2021 by the University of Chicago’s Polsky Center and the Chicago Quantum Exchange, was built around the observation that quantum teams need mentors who understand decoherence times and equipment that a startup cannot afford, and that a general accelerator supplies neither. Canada’s Creative Destruction Lab runs a quantum stream on similar reasoning. The US National Science Foundation’s I-Corps program pushes teams through customer discovery, meaning structured interviews with potential buyers before the product is fixed.
What these programs do best is force a technically excellent team to find out which real problem their device solves. That process reliably surprises people. Sensing teams that assumed defence discover that the paying market is utilities. Software teams that assumed finance discover that the near-term buyer is a chemicals company with a specific simulation problem and an existing budget for it.
So when a team tells you they interviewed sixty potential customers and changed their target market as a result, they have told you something concrete about how they make decisions. When a team tells you the applications are obvious, they usually mean they haven’t asked. We consider that one of the more reliable early signals available, and it costs nothing to check.
Six questions worth asking
Which revenue line, and how much of it is grant money? Split the number before comparing it to anything.
Physical or logical? Any qubit count, fidelity figure or roadmap milestone should specify which, and should say whether the result was demonstrated or announced.
Who signed the pilot, and from which budget? Named line-of-business owner and operating budget, or innovation team and discretionary spend.
What is the incumbent method, and what does it cost the customer today? Quantum advantage is a comparison, and the comparison needs a baseline the customer already believes.
Where is the intellectual property, and on what terms? For a spinout, exclusivity, field of use, institutional equity, and the ongoing engagement of the named inventors.
What did they learn that changed their plan? Teams that have done customer discovery can answer this in specifics. Teams that haven’t will answer with vision.
What this asks of the people doing the diligence
None of these checks are exotic. What they require is enough technical fluency to know when an answer is evasive, and that is the constraint most investment teams actually face in this sector. The difference between a physical and a logical qubit, between an announced roadmap and a demonstrated result, between quantum key distribution and post-quantum cryptography, between annealing and gate-based computation: none of it is difficult, and all of it decides whether a pitch deck is describing a product or a research programme.
Reflexive dismissal and reflexive enthusiasm are the same error wearing different clothes. Both substitute a prior for a check. The teams that are building real businesses in this field can answer every question above, and they generally prefer to be asked.
Quantum Academy’s certification programs are built to give non-specialist professionals exactly that fluency, working from the vocabulary and the claim structures that show up in vendor material and investment memoranda. You can review the current programs, formats and pricing at quantumacademy.com/. For deeper technical background on the underlying claims discussed here, PostQuantum.com covers the same territory at greater length.