DOE's Quantum Computing Roadmap of 25 September 2026 Sets 50 to 100+ Logical Qubits for 2028: How the SCAC Report's Targets Compare With the Genesis Q Competition
On 30 May 2022 the 59th TOP500 list reported that Frontier, at Oak Ridge National Laboratory, had scored 1.102 exaflop/s on the High-Performance Linpack benchmark. The Department of Energy had spent years on its exascale program, and the result could be checked against a yardstick the field had used for decades: a public benchmark and a public list.
Quantum computing has no equivalent benchmark with that standing. On 25 September 2026 DOE's Office of Science published the report of its advisory committee's quantum subcommittee, and the announcement republished by Fermilab describes its charge as a roadmap "toward demonstrating a scientifically relevant, error-corrected quantum computer by 2028." That sentence is likely to be quoted often over the next two years. What the report itself counts as reaching it is the question this post tests.
Practical takeaway. The SCAC report attaches numbers to 2028, roughly 50 to 100+ logical qubits and 104 to 105 hard logical operations, but labels them illustrative. The Genesis Q Competition (DE-FOA-0003657) uses similar operation counts but sets at least 100 logical qubits in its published sample agreement, and a DOE verification team checks negotiated milestones at an evaluation date currently anticipated for September 2028. Anyone citing "error-corrected by 2028" should say which of the two documents they mean.
What DOE claimed on 25 September 2026, and who wrote the report
The report is titled SCAC Quantum Committee Report: A Pathway to an Integrated Quantum Future and is dated September 2026. The subcommittee was chaired by Anna Grassellino, chief technology officer of Fermilab, with Supratik Guha of the University of Chicago's Pritzker School of Molecular Engineering as vice chair. According to its own appendix, it was set up in April 2026, issued a public request for information in May, held a community town hall on 12 June, received about 50 written responses, and presented its final report to the full advisory committee in July.
The announcement by Darío Gil summarizes the result as a three-phase plan. Phase I runs Quantum Grand Challenges from 2026 to 2028. Phase II establishes a DOE Quantum Computing User Facility, described as "an open, collaborative scientific instrument" and explicitly distinguished from a commercial "black box" cloud service. Phase III, from 2030 on, integrates quantum processors with DOE's supercomputers, AI systems and experimental facilities. The announcement also states that success "must be measured by scientific utility."
What the report's 2028 column says: 50 to 100+ logical qubits and 104 to 105 hard operations
The text of the report is careful. Its hardware section says the roadmap "defines illustrative performance ranges" and avoids prescribing any technology. The ranges themselves appear in a two-column figure. For "2028 scientific-relevant demonstrations" it lists about 50 to 100+ logical qubits; 104 to 105 hard logical operations; a full scientific calculation returned within 24 hours; effective connectivity within about twice the ideal across a machine of about 1,000 qubits; fault-tolerant execution reliable enough for an end-to-end scientific calculation; a combined quantum, supercomputer and AI workflow; and "independent validation against experiment, classical limits, or predictive scientific value."
The second column, "2030+ user facility capability," jumps to 1,000 to 10,000 logical qubits and 109 to 1010 hard operations, with campaigns finishing in days and the machine computing more than 80 percent of the time.
Two features of the 2028 column deserve attention. The lower bound of 50 logical qubits and 104 operations is a modest machine by the report's own standards; a system at that floor would sit a factor of ten below the upper bound on operations. And the column states no target logical error rate. The only error-rate figure in the report, 10-7 to 10-9, appears in the appendix summarizing stakeholder interviews, as one of the readiness criteria interviewees suggested before DOE builds a user facility. The same appendix reports that stakeholders "differed significantly" on pace: some expected scientifically useful fault-tolerant systems by 2028, while others thought large-scale fault tolerance would take substantially longer.
How the Genesis Q Competition of 17 September 2026 defines the same target
DOE had already put money behind the 2028 date eight days earlier. The request for applications for the Quantum Genesis Q Competition, DE-FOA-0003657, defines a logical qubit, defines a hard operation as one that is difficult in the chosen error-correction scheme (for example a non-Clifford gate in stabilizer codes), and lists illustrative values for a first-generation "scientifically relevant quantum computer" of 100 logical qubits and 105 hard operations, capable of running a full scientific workflow. The request says milestones may differ by performer and are negotiated, and architectures that do not fit the logical-qubit count may propose equivalent metrics. The published sample agreement is more specific. Its first-generation milestone asks for circuits on at least 100 logical qubits, specified by DOE's verification and validation team, standardized across awardees, and sized so that on today's reference architecture they would compile to approximately 104 to 105 hard operations; the team assesses the correctness of the outputs. Those milestones are negotiated into an Other Transaction Agreement, verification staff may operate each machine through cloud or physical access, and the common evaluation date falls in the eighth quarter, which DOE currently anticipates in September 2028 and may extend. We read the payment terms and the verification circuits in detail in our post on the Genesis Q sample agreement.
Put side by side, the two documents use the same vocabulary. Both speak of logical qubits and hard operations, and both treat scientific relevance as the third metric. The sample agreement's operation range, 104 to 105, is the same as the advisory report's. The visible difference is the logical-qubit floor: the report's range starts near 50, while the sample agreement asks for at least 100. The larger difference is procedural. The report offers ranges for planning, while the competition attaches similar numbers to standardized circuits, a verification team and a payment decision. The report maps its Quantum Challenges Phase onto the "DOE Q Competition" in a table aligning it with Quantum Genesis, so the two are meant to work together.
What the roadmap asks of industry: hardware access, shared prototypes and lower IP barriers
The report's most consequential passages may be the ones about access. Its third recommendation asks that machines in the challenges let researchers "understand, characterize, and optimize the underlying hardware rather than treating it as an opaque black box," with access to control interfaces, diagnostics and performance characteristics. It proposes embedded co-design residencies of 6 to 18 months, joint appointments, shared pre-competitive prototypes, milestone-based co-investment tied to scientific outcomes, and a community working group on interconnect standards, cost sharing and intellectual property for joint development.
The recommendation is titled "Establish New Partnership Models for Co-Design and Lower IP Barriers," which can sound like a call for open-source quantum hardware. The text is narrower. It states that "appropriate intellectual property protections remain essential" and asks DOE and its partners for collaboration frameworks that "maximize scientific openness while respecting commercial innovation." The working group's IP mandate covers "any joint development," and the report says nothing about opening up what a vendor builds on its own, such as its control stack or error-correction methods. In Quentir's reading the arrangement resembles DOE's existing user facilities such as Frontier: researchers win machine time through merit review, the laboratory knows its machine well enough to verify what it delivers, and the vendors keep their product designs. The report's appendix records that stakeholders still differ on the balance between open scientific infrastructure, public-private partnerships and support for proprietary technology development, so the terms remain to be written.
For vendors this is a real negotiation. Diagnostic access is what lets a national laboratory check a claim independently, and it is also what a company guards most closely. Quentir expects that access to be written into award terms for machines in the challenges, scoped to what verification needs and covered by confidentiality agreements. Governments are settling access questions in different settings. In France this week the defense innovation agency announced ALIQUANTE, a five-year classified quantum programming platform built by Thales, Bull and CEA, which places programming work for military users inside a classified environment. That concerns who may use the software and for what, a separate matter from hardware transparency, but it shows how quickly quantum access becomes a question of institutional design.
How Quentir Reads It
The claim "error-corrected quantum computer by 2028" is supported by the report, but only as a range with a low floor, no logical error-rate target, and an explicit statement from stakeholders that the date is contested. The instrument better placed to settle the question is the Genesis Q agreement, where a DOE verification team runs standardized circuits and checks the outputs against milestones negotiated with each awardee. Quentir expects the phrase to travel much further than the figure, and that is where readers should be careful: a company announcing in 2028 that it has "met DOE's roadmap" may be pointing at the bottom of an illustrative range, while the competition's evaluation applies its own negotiated milestones.
The stakes reach beyond the laboratories. The report promises drug discovery, catalysts for cleaner manufacturing and fusion materials, and its sixth recommendation asks Congress for resources "beyond those currently available." Public money for science is easiest to defend when the public can see what was promised and whether it was delivered, as it could with Frontier. The report's own suggestion of objective readiness criteria before a facility is built would give the 2028 date that kind of yardstick, and the planning activity it recommends is the place to write them down.
Quentir's Signature Brief editions give a fixed scope, dated sources and a license for internal use for readers who follow these instruments closely. The next date to watch is 19 October 2026, when Genesis Q applications close.
Sources: US Department of Energy, Office of Science Advisory Committee (SCAC), "SCAC Quantum Committee Report: A Pathway to an Integrated Quantum Future" (September 2026; chair Anna Grassellino, vice chair Supratik Guha), for the hardware capability figure (2028 and 2030+ columns), the hardware and technology roadmaps, the user-facility planning questions, Recommendations 1 to 6, the Quantum Genesis alignment table, Appendix B (timeline), Appendix C (stakeholder interviews, including the 10-7 to 10-9 suggestion) and the areas of divergence; Darío Gil, "DOE releases national quantum computing roadmap following field-wide effort led by SCAC subcommittee" (originally published by the DOE Office of Science, republished by Fermilab, 25 September 2026); US Department of Energy, DE-FOA-0003657, "The DOE Quantum Genesis Q Competition", request for applications, Version 1.0 (17 September 2026; applications due 19 October 2026), Sections IV.A and IV.B and Table 1, and its sample Other Transaction Agreement (Milestones 3a to 3c and Article 6.1.5); TOP500, June 2022 list (Frontier, 1.102 exaflop/s HPL); Quentir, Genesis Q sample agreement (18 September 2026); Quentir Defense Monitor, ALIQUANTE (24 September 2026).
Published intelligence, built to inform your own decisions. Published: September 27, 2026.