DOE Wants a Useful Error-Corrected Quantum Computer by 2028

The U.S. Department of Energy now has a written plan for turning quantum computers into scientific instruments, and it comes with a deadline. On September 25, DOE released 'Path to an Integrated Quantum Future,' a roadmap from the Office of Science Advisory Committee's quantum subcommittee, according to Fermilab. Its core target: demonstrate a scientifically relevant, error-corrected quantum computing capability by 2028, then use that proof to justify a DOE Quantum Computing User Facility open to researchers nationwide.

The subcommittee was chaired by Anna Grassellino, Fermilab's chief technology officer, with Supratik Guha of the University of Chicago's Pritzker School of Molecular Engineering as vice-chair, per Fermilab. The report says it drew on stakeholder interviews, a public request for community input, a town hall and deliberations with industry, academia, national labs, federal agencies and international groups.

Science first, qubits second

The most consequential choice in the document is what it refuses to measure. Rather than setting targets for physical qubit counts or gate fidelities alone, the committee proposes judging progress by scientific utility: whether a fault-tolerant machine can answer a question classical computers cannot.

The report lists the milestones it thinks could establish that utility by 2028. They include predicting the electronic structure of catalytically important molecules, modeling strongly correlated materials tied to superconductivity, validated simulations of fusion-relevant materials and plasma processes, realistic neutrino and nuclear interaction calculations, and reproducing key phenomena from the quantum field theories behind particle physics. Past 2028, the hardware roadmap points toward approximately 1,000 logical qubits and sustained fault-tolerant operation, the scale the committee says a full user facility would need.

That is a pointed stance at a moment when vendors compete on headline numbers. Just this week Infleqtion announced 30 entangled logical qubits on a commercial neutral-atom system. DOE's advisers are saying, in effect, that those milestones count only if they turn into chemistry, materials or physics results someone can check.

Three phases, one facility

The roadmap is organized in three stages. Phase I, running 2026 to 2028, is a set of Quantum Grand Challenges: milestone-driven competitions pairing national labs, universities and companies to co-design hardware, algorithms and software around specific science problems. The report says industry should be encouraged to compete across hardware types, and that DOE should consider giving teams early access to existing or near-ready machines.

Phase II is the Quantum Computing User Facility itself. Fermilab describes it as an open scientific instrument rather than a commercial cloud service, the same model DOE uses for its light sources, neutron sources and leadership-class supercomputers. The committee recommends a planning effort run in parallel with the challenges to assess readiness, user demand and facility designs, and it explicitly conditions the facility on demonstrated scientific utility.

Phase III, from 2030 onward, envisions quantum co-processors, simulators and sensors embedded throughout DOE science and wired into its AI and high-performance computing networks.

The roadmap is deliberately technology-neutral. Superconducting circuits, neutral atoms, trapped ions, photonics and spin qubits are all in play, per Fermilab, and the report calls for open interconnect standards so hybrid classical-quantum systems are not locked to one vendor.

The part companies will read closely

Recommendation 3 is where the plan gets practical, and potentially contentious. The committee wants 6-to-18-month residencies that embed domain scientists inside hardware companies' engineering teams and industry staff inside national labs, along with joint appointments, shared prototypes and milestone-based DOE-industry co-investment. It also asks that machines in the challenge program let researchers see inside them, including architecture, control interfaces and diagnostics, rather than treating hardware as a black box. It proposes a working group to settle intellectual property handling and cost-sharing.

For quantum startups, that is both an opportunity and a demand. Federal co-investment tied to scientific outcomes is real money in a sector that still struggles to show revenue. But opening control stacks to lab scientists cuts against the proprietary posture most hardware firms have maintained.

What the report does not say

There is no price tag. Recommendation 6 asks DOE and Congress for broadened investment commensurate with the opportunity and says realizing the vision will require resources beyond those currently available. It compares a future facility in ambition to previous landmark DOE initiatives, but leaves the number to appropriators.

The report is also candid about disagreement. In a section on divergent views, it notes that some stakeholders expect scientifically useful fault-tolerant systems by 2028 while others think large-scale fault tolerance needs substantially longer. Interviewees split on whether the facility should be one national center or a distributed network, and on how much DOE should host frontier hardware versus integrate other people's. One reference in the report puts the difficulty in perspective: a cited study estimates a single particle-scattering simulation would need roughly 4 million physical qubits under standard surface-code assumptions.

Why it matters

The roadmap slots under the Quantum Genesis Initiative and the administration's quantum executive order, both of which the report says arrived while it was being drafted. Those set the policy intent. This document supplies the operating plan: who competes, what counts as winning, and what gets built afterward.

The shift in metric is the real news. If DOE adopts it, the government's largest scientific customer will reward the quantum company that computes a catalyst or a fusion material nobody could compute before, not the one with the biggest qubit count on a slide. And with a 2028 deadline attached, that verdict arrives within two years.