Three years. No gate threshold. No qubit count. No definition of scientifically relevant. The Department of Energy wants a fault-tolerant quantum computer by 2028 — and has not published a single technical requirement.
image from grok
The US Department of Energy announced a goal to deliver the "first generation of fault-tolerant quantum computers" by 2028, but provided no concrete procurement specifications—no qubit count, gate threshold, or definition of "scientifically relevant" calculations. A concurrent Caltech preprint using high-rate quantum LDPC codes estimated fault tolerance might require only 10,000–20,000 physical qubits, significantly lower than surface code estimates, though this remains theoretical. Current industry milestones from IBM and Google represent incremental progress toward, rather than attainment of, fault tolerance, which remains architecturally undefined and talent-constrained.
The US Department of Energy wants a fault-tolerant quantum computer by 2028. It has not said what that means.
On March 27, at the inaugural meeting of the Office of Science Advisory Committee, DoE undersecretary Darío Gil announced that the department would "deliver the first generation of fault tolerant quantum computers capable of scientifically relevant quantum calculations" within three years. The announcement made headlines. It also contained no procurement specification: no gate threshold, no qubit count, no definition of what "scientifically relevant" actually requires.
One day later, a Caltech team posted a paper on arXiv that landed in the same news cycle with a very different kind of number. Using high-rate quantum low-density parity-check codes, the group estimated that a fully realized fault-tolerant quantum computer might need as few as 10,000 to 20,000 physical qubits — not the millions that surface code estimates often imply. Each logical qubit could be encoded with roughly five physical qubits under this scheme. That theoretical result does not close the gap between Gil's announcement and an actual procurement document. But it changed the background noise in the room.
"The goal is very optimistic but worthy," Yale physicist Steven Girvin told Science. That is the most honest summary anyone has offered so far.
The architecture question is not settled. The DoE's approach is deliberately agnostic — it wants commercial providers to propose solutions, which means superconducting qubits, neutral atoms, trapped ions, or photonics are all in scope. IBM's Nighthawk systems, based on superconducting transmon qubits, will run circuits with 15,000 two-qubit gates by 2028 across more than 1,000 connected qubits — a milestone IBM announced as a step toward large-scale fault-tolerant quantum computing. It is a step, not a destination. IBM's own roadmap puts true fault tolerance with Starling, targeting 100 million gates on 200 logical qubits, at 2029 at the earliest. Google Quantum AI achieved a landmark quantum error correction proof in 2024, demonstrating that error correction works in practice on superconducting qubits, not just in theory. These are genuine advances. None of them constitutes a fault-tolerant quantum computer.
The talent gap compounds the timeline problem. A 2025 report from quantum error correction firm Riverlane estimated that only 600 to 700 professionals specialize in quantum error correction worldwide, against a requirement of 5,000 to 16,000 by 2030. The field published 120 peer-reviewed QEC papers in the first ten months of 2025 alone, up from 36 in all of 2024. The research pipeline is accelerating. The graduate cohort that can execute on it is not.
DoE has more than doubled its quantum information science investment since 2019, to over $1.6 billion, and its Office of Science annual budget stands at $8.4 billion — the United States' single largest funder of the physical sciences. The proposed fault-tolerant machine would sit inside a quantum user facility, likely at one of the Office of Science's ten national laboratories. Where the money goes determines who can bid. But the specification that would let anyone bid has not been written.
Global competition adds urgency and noise in equal measure. Government funding for quantum computing has reached approximately $50 billion worldwide, with Japan allocating $7.9 billion in 2025 and the US $7.7 billion. These are not comparable programs — they fund different things at different stages — but the headline numbers create political pressure to be first.
The DoE announcement is a statement of intent, not a procurement contract. It tells the quantum industry that the US government wants to buy something, without specifying what, from whom, or how it will be tested. That is not unusual for early-stage government quantum programs. It is also not a technical document. The researchers who would need to build this machine have a clearer picture today of what is theoretically required than they did two weeks ago. They also have a clearer picture of how far they remain from being able to deliver it. Gil set a deadline. The engineering spec sheet is still mostly blank.
Story entered the newsroom
Research completed — 7 sources registered. Three events converge: (1) DOE undersecretary Darío Gil pledged at SCAC March 27 2026 to deliver a fault-tolerant quantum computer by 2028 as a user f
Draft (677 words)
Approved for publication
Published (662 words)
Get the best frontier systems analysis delivered weekly. No spam, no fluff.
Quantum Computing · 4h 16m ago · 4 min read
Quantum Computing · 5h 46m ago · 3 min read
