We all want Quantum Benchmarks until our favorite company isn’t on "The List"
DARPA did not advance Rigetti to Stage B of the Quantum Benchmarking Initiative. Commerce bought $100M of equity in it anyway.
In November, DARPA looked at Rigetti’s fault tolerance roadmap and did not advance it to Stage B of the Quantum Benchmarking Initiative. Six months later, the Commerce Department signed a letter of intent for an equity stake of up to $100 million in the same company.
Every time a list like this drops, an LP messages me asking what the government knows that they don’t. Why isn’t my modality in HARQ? Why did DARPA pass on a company Commerce just wrote into a $2 billion package? The assumption here is there is one leaderboard, and the government is the final arbiter of success.
Disclosure: Not investment advice, not an offer or solicitation of anything.
Quantum has never had a neutral referee
There isn’t one leaderboard, and there never has been. Every scoreboard this industry has built has some sort of incentive behind it.
IBM introduced Quantum Volume in 2019 as a single number capturing width and depth together, and for a few years it was the closest thing to a shared metric. Then Honeywell and Quantinuum started setting the records on trapped ions, doubling QV on a schedule, and IBM’s own messaging moved on to layer fidelity and speed metrics that suit its roadmap better. IonQ answered with Algorithmic Qubits, a benchmark of its own design that its own machines lead. Each metric is defensible in isolation, but each also happens to flatter the architecture of the company that publishes it.
The stats come from the company’s own press release, often cherry picked, sometimes with access cut off for researchers who published less than favorable numbers (and a few lawsuit threats…). The neutral attempt exists on paper: the IEEE P7131 benchmarking working group, chartered around 2021, has yet to publish a standard and has infighting because every company wants THEIR benchmark to win.
AI runs the same playbook at a thousand times the scale (benchmarking on different types of exams, lab eval tables, vendor configured MLPerf submissions, tokens-per-second vs cost per token vs watts per token each flattering a different chip. But AI has users, and a few hundred million people means it has to exist and scale and people feel it. Example: I spent a lot on Fable tokens to switch back to Opus and Codex for programming. Quantum has little correction: a few hundred machines, R&D partners, nobody that could or would contradict it (being into quantum is also hot now even if you don’t have results — waiting for the Allbirds pivot to quantum too!) Procurement is the closest thing to a preference there is, which is why the government lists carry weight.
So yeah, we all want benchmarking, until our favorite company is not on the list.
DARPA vets roadmaps. Commerce secures supply chains.
DARPA’s Quantum Benchmarking Initiative is technical vetting due diligence and from conversations with founders in the running, an intense one, with deep experts brought in (definitely more intense than many VCs). Stage A, announced April 3, 2025, took 18 companies through a six month sprint to defend a utility scale roadmap. Stage B, announced November 6, 2025, promoted eleven: Atom Computing, Diraq, IBM, IonQ, Nord Quantique, Photonic Inc., Quantinuum, Quantum Motion, QuEra, Silicon Quantum Computing, and Xanadu. Stage B runs about a year, which puts Stage C decisions around late 2026.
Why aren’t some companies on the list? Some because they are technically already in Stage C. Microsoft and PsiQuantum advanced to the final phase of QBI’s predecessor program in February 2025, which DARPA treats as equivalent to Stage C. So, the topological and photonic machines were the first to go into this tier, which is not where the public markets are pricing them. And, some of the others were acquired: Atlantic Quantum into Google, Oxford Ionics into IonQ.
Commerce seems to be looking at something different. The May 21, 2026 letters of intent total $2.013 billion across nine companies, with minority equity stakes as a condition of the funds. Some companies declined on those terms. The bigger part of the award went to foundries, not directly to quantum computers. IBM gets $1 billion toward Anderon, its standalone quantum wafer subsidiary in Albany, matched by $1 billion of IBM’s own. GlobalFoundries gets $375 million for a domestic quantum foundry serving multiple modalities. Then $100 million each to Atom Computing, D-Wave, Infleqtion, PsiQuantum, Quantinuum, and Rigetti, and $38 million to Diraq.
DARPA is looking at the roadmap and Commerce is looking at supply chain risk: will the manufacturing base exist on American soil.
Where the lists disagree
DARPA gave no public reasons for its cuts, and the miss means different things at different companies. Many expect to qualify later (is that a forward looking statement?). Google doesn’t need DARPA’s money or its validation, though there always seems to be a little bump in price during some big quantum news, and its Stage A acquisition, Atlantic Quantum, was absorbed into Google Quantum AI in October. Rigetti is the more interesting case. It is a pure-play public company that is building quantum processors and selling them. The QBI did not move forward, while Commerce took equity anyway. Perhaps it’s buying Fab-1 in Fremont, one of the few operating superconducting fabs in the country. On the supply chain side, that asset counts no matter where the specific company roadmap is.
On the other side is IonQ. It passed DARPA, acquired Oxford Ionics for $1.075 billion to add more trapped-ion talent and work through additional trapped ion approaches. But it’s not in Commerce’s $2B investment, where trapped ion appears only through Quantinuum. The likeliest explanation that analysts are giving is that IonQ is buying SkyWater for roughly $1.8 billion and has a great cash balance right now, and declined the terms.
Silicon spin holds four of the eleven Stage B slots (Diraq, Photonic Inc., Quantum Motion, Silicon Quantum Computing), the heaviest representation of any qubit type. Commerce gave Diraq $38 million. DARPA, betting on physics, is heavy on silicon and their potential scaling. The supply chain money are betting less here, maybe because the foundries already exist: Intel and GlobalFoundries.
The scorekeepers are also buyers
The lists can disagree like this because neither agency is a referee. A referee has no position in the game. In this sector, every referee is also a customer, an investor, or both.
QBI’s money comes at Stage C and reportedly tops out around $300 million for companies that finish; Stage A and Stage B were just evaluation, not funding. Commerce’s checks carry an equity stake and terms, which is why some companies passed. For a well-capitalized company the cash is not the draw, and several didn’t bother to apply. Their absence says nothing about their physics.
At the other end of the funding cycle, the government money can de-risk an early stage startup. A quantum startup here can run for years on DARPA contracts, AFRL work, and SBIR and STTR awards before it ever takes a venture capital check. By the time it does raise, the round is more like a Series A, not a seed.
A spot on a government list is a signal that goes straight into the next deck and the next raise. Being cut is a bigger problem, especially when the competition is so public. For a private company it is a harder conversation with the next investor. For a public one it can move the stock. A single Jensen Huang sentence wiped billions off market caps in an afternoon. Founders were nervous about exactly this when QBI launched.
DARPA is now funding something that might imply QBI asked the wrong question
In April, DARPA launched HARQ, Heterogeneous Architectures for Quantum: 19 teams from 15 organizations working on combining different qubit types into single systems, split between a cross-qubit compiler workstream and a quantum-interconnect workstream. Program manager Justin Cohen said the field should move off the “one qubit to rule them all” mindset. The quantum networking people and academics have said for years that, in the near term, these systems will all need to work together. A January preprint on heterogeneous superconducting-plus-neutral-atom architectures reports a 752x average speedup over neutral atoms alone (but no company is yet cross-modality).
For HARQ, the compiler side is Infleqtion, memQ, and Q-CTRL alongside university groups at Michigan, Penn, Illinois, Maryland, and UT Austin, with memQ pulling qBraid, MIT, Yale, and Chicago onto its team. The hardware side, the shared backbone, is photonic interconnect, and IonQ is building it while separately demonstrating the first entanglement link between two independent trapped-ion computers with the Air Force Research Laboratory. The modalities being stitched together, ions and neutral atoms and superconducting and photonic, are the same four everyone already funds, so paving the way for more modular systems. So there will not be quantum-only data centers. The syndrome decoding, the calibration loops, the error correction cycles that keep logical qubits alive are all classical computation, running on classical hardware that has to sit physically next to the QPU. NVIDIA's Ising Decoding model runs on Grace Hopper GPUs at microsecond latency. Quantinuum's Helios routes its real-time decoding through NVIDIA hardware over NVQLink. The quantum machine cannot operate without the classical envelope around it.
The program is small next to QBI, runs 24 months, and absence mostly means a company probably didn’t apply. The bigger issue for startups can be overexposure to a single government program with an election cycle putting holds on spending. The National Quantum Initiative Act that funded a lot of programs was signed in December 2018 under Trump, and its funding authorization expired on September 30, 2023. It then lapsed for two and a half years. The reauthorization cleared the Senate Commerce Committee unanimously on April 14, extends the initiative to 2034. A national program that can expire for thirty months is a risk when much of the revenue in the field depends on it. The state is this sector’s biggest customer, investor, and referee, and politics are always involved.
Europe has always run on state money, so what?
If this looked like an American idiosyncrasy, recently we saw some big plays out of Europe. On July 9, Munich-based QuantumDiamonds announced a €91 million round: €15 million in equity led by World Fund, and €76 million in non-dilutive funding approved at the EU level under the European Chips Act, provided jointly by the German federal government and Bavaria.
QuantumDiamonds is the first startup ever approved for manufacturing funding under the EU Chips Act, a category previously reserved for established semiconductor names like GlobalFoundries and STMicroelectronics. And it is not a quantum computer company. It sells nitrogen-vacancy diamond sensing systems that image current flow inside advanced chip packages, with nine of the world’s ten largest semiconductor manufacturers engaged and systems installed at Eurofins EAG in Sunnyvale and iST in Hsinchu.
Some could shrug. Europe has always run on state money. But that state money was research money: Horizon grants, EIC awards, national quantum programs, checks that fund papers and prototypes, but rarely have the same VC/LP timelines and return expectations attached. This is the same pattern I have been writing about since the supply chain piece: the development costs for scaling quantum manufacturing need to involve government to close the supply chain gap, so sovereign capital does it.
The manufacturing signal is still lurking
So do any of these awards mean anything? Yes, but also no. A leaderboard assumes everyone is playing and the scoring is neutral. We don’t always know who applied or who was turned down, and since we do not have an objective, stacked ranking, absence is not a verdict and presence is not proof. Self selection and strings add nuance to the list, as much as physics does. Not doing a SPAC, btw, is not because a company is failing; half the sector went public in eighteen months, and staying private can just mean not needing the cash and actually being in a good position, potentially. IonQ skipped Commerce, while moving ahead on DARPA. Rigetti missed DARPA and got Commerce. Microsoft holds DARPA’s most advanced tier on a Majorana result physicists still dispute and some are veryyyy angry about. But the same problem will apply to buyers of the quantum hardware now. They are making modality choices on the physics and supply chain.
In the next phase of quantum, it will be 1) who can get the best results into the hands of customers, 2) who can manufacture at volume, 3) who can slot into the data center as it already stands, and 4) who can deliver, deploy, and upgrade quickly. I made this argument in the consolidation piece: a “bad” quantum computer still gets sold, because the customer can’t wait two years for a better one because they need to get ahead now. But still, watch Stage C in late 2026.
AI compute is facing this right now. An alt compute chip can beat the NVIDIA GPU on every benchmark and still lose because it can’t get HBM in a market where OpenAI alone signed for up to 900,000 DRAM wafers a month from Samsung and SK Hynix, and where the three memory makers who control roughly 90% of DRAM were just sued for allegedly using the HBM pivot as cover for a coordinated supply squeeze (allegations they deny, though Samsung and Hynix did plead guilty to DRAM price-fixing in the 2000s). Or the chip loses because it asks the data center to rebuild its cooling architecture around it. Doesn’t mean the stock isn’t going to be hyped, though.