Ask most scientists what government is for and the answer is narrow: it is where the grant comes from. You write the proposal, you hire the students, they do the work, you present the results, you write the next proposal. For decades that was the entire shape of the relationship, and for a field as early as quantum computing it was a reasonable shape.
Nate Gemelke, co-founder and Chief Technology Strategist at QuEra, made the case at a recent QuEra Quantum Alliance gathering that this is now the least interesting part of the picture. Not because sponsored research stopped mattering, but because everything that comes after it is arriving sooner than the industry planned for, and almost nobody is organized around it.
This post draws on Nate Gemelke's talk at the QuEra Quantum Alliance 2026 Annual Meeting (QQA 2026) and the discussion that followed, which included Robert Ledoux, Director of the Quantum New Mexico Institute.
The Arc Government Actually Travels
Gemelke sketched government's role as a progression rather than a job description.
It starts with the funder. That is the familiar one, the sponsored research era, and it is the stage the quantum industry has spent most of its life inside.
Next comes something less obvious. A company selling a technology that it openly admits is not yet useful has an awkward question to answer about who its customers should be.
"As a business, it's important to us that we have customers, but when you are producing something that you freely acknowledge is not yet useful, who should those customers be? Government is oftentimes the first customer," said Nate Gemelke, co-founder and Chief Technology Strategist at QuEra Computing.
The reasons stack up. A disruptive technology is something government has an institutional interest in staying ahead of. Government also wants a healthy ecosystem to exist at all, which makes early purchasing a form of ecosystem building. And government organizations tend to employ some of the most technically capable people available to evaluate what they have bought.
That evaluation is the third role, and it is a real service. An independent technical assessment from people with no commercial stake in the answer is worth more to a vendor than another friendly benchmark.
The fourth role is the one vendors are usually expected to avoid discussing. At some point there will be regulatory structure and a security posture around quantum computing, and Gemelke's view is that this is appropriate rather than unfortunate. Quantum computing is, in his framing, a genuinely dangerous quantity in at least one dimension, given the cryptographic relevance of what these machines will eventually do.
Then comes the reframe. Gemelke put strategic partner at the far right of the progression, as the final stage, and then immediately corrected himself in front of the room: government is a strategic partner across the entire arc, not a status it graduates into once the technology matures. Treating the relationship as a grant pipeline that eventually turns into a procurement pipeline misreads what has actually been happening the whole time.
The Programs That Built the Company
That argument is easier to accept with the receipts, and QuEra's own history supplies them.
The very first thing the company did, before it was meaningfully a company, was write a proposal to a DARPA program called ONISQ. It was a five-year exploratory effort from a more hopeful period, aimed at finding tractable optimization problems for noisy intermediate-scale quantum computers. ONISQ was conceived before QuEra existed. By the time it started, QuEra was four people down the street from Harvard, and by Gemelke's account the program was the company's largest single source of funding through its first months. The work, done together with Harvard colleagues, centered on the maximum independent set problem, the one graph problem in that era where a single qubit could stand in for a single node.
What DARPA did next mattered more than the money. Rather than scaling up the hardware effort, the agency created a set of small advanced research concept grants designed to fund roughly a single postdoc at a university to explore one application space. They were short and inexpensive. They also, Gemelke said, nucleated a set of early customer relationships and produced something like a demand signal at a moment when almost nobody was using a quantum computer of any kind for anything.
From there the list widens. DARPA's science office ran further work on signaling, control, and error correction at scale. IARPA launched its Entangled Logical Qubits program around 2023, which is the point at which the intelligence community, having previously concentrated on superconducting qubits, formally treated neutral atoms as a modality worth funding. That turned into a lasting relationship. NSF, whose mandate is building research collaborations rather than buying hardware, stood up its National Quantum Virtual Laboratory centers, and QuEra is partnered with several of them, including centers at MIT and the University of New Mexico.
None of those programs bought a product. All of them shaped what the company became.
A Colonoscopy of Quantum Computing
The clearest sign that the relationship is shifting is where the programs live inside the agencies. DARPA's quantum work moving from its basic science office to the office that handles more applied technology was, Gemelke argued, a signal of maturity in itself.
The Quantum Benchmarking Initiative is the sharpest example. People in the field have taken to calling it a colonoscopy of quantum computing, and the description is affectionate rather than bitter. QuEra hosted a site visit from roughly forty national lab and FFRDC scientists whose assignment was to look hard at neutral-atom technology and judge whether it can actually reach utility. QuEra is in Stage B, having cleared the Stage A feasibility study, and submitted a full research agenda describing how it intends to get to useful computation. Gemelke described the program's read as positive so far, with the evaluators not having landed on a fundamental flaw in the technology. At the time of the talk, the hope was to transition to Stage C, without a firm date attached.
What makes QBI different from a research grant is that it is partly a procurement program. The question underneath it is not only whether the technology works. It is whether it is time to start buying it.
Government Funds Hardware. It Barely Funds Applications.
Here is where the room got interesting, and where the complaint surfaced.
"It was always surprising to me how little interest the government seemed to have in formally funding it. They like having, 'Here's a thing. Go build it.' But applications, which is the top of the chain, there's very little that supports it," said Robert Ledoux, speaking in the same session.
The asymmetry is structural rather than ideological. A hardware milestone is legible. It has a specification, a delivery date, and a demonstration at the end. Applications research has none of that, which makes it hard to write into a solicitation even when the people writing the solicitation agree it is the more important open problem. Ledoux's read is that the field has not said clearly or often enough that it wants to work in application space and that government has a role to play there. Several people in the room agreed that applications, not hardware, is now the harder unsolved question.
Use Cases, Not Applications
One attendee pushed the argument a step further, and the distinction is worth carrying forward.
Applications is a scientist's or an engineer's word. It describes a category of problem from the outside. Use case is the end user's word, and it starts from a specific person with a specific problem. Starting from the use case forces a more productive engineering conversation, because it makes you answer what utility is actually being offered and where that utility rolls off compared with whatever the user does today. Two customers with the same nominal application frequently have entirely different use cases.
The follow-on point was about how badly this conversation usually goes. End users often articulate their own use case poorly, and drawing it out requires iteration and enough trust that nobody feels tested. Gemelke added a third word to the pile, primitives, and noted that the real skill is layering: bringing the right level of abstraction to the right group of people at the right moment. The industry, by his own admission, could do this considerably better than it currently does.
Why the Timing Argument No Longer Works
For most of QuEra's existence, the honest answer about when a quantum computer would be big enough to do something useful was about ten years. Everyone in the field knows how that game is played, and everyone in government had learned to discount it accordingly.
Gemelke's point is that the answer has changed, and that the industry's messaging has not caught up with it. Progress on deployable algorithms, error-correcting codes, and hardware scale has been converging faster than expected, and the intersection point is close enough now to plan around. QuEra has committed publicly to delivering Libra, a fault-tolerant system, in 2028. That commitment makes every downstream question, including who the first customers are, who evaluates the technology, how it gets acquired, and what it gets pointed at, an immediate question rather than a distant one.
Which is the real argument for treating government as a partner rather than a funding source. The funder stage is nearly over. The stages that follow will determine whether this technology gets used well, and those stages require a relationship that neither side has fully built yet.




.webp)




