"Originally, the idea of a logical qubit was that, once your physical qubits were better than a certain threshold, then you should be able to correct all errors in the qubit, which basically means that your qubit would be everlasting."
That is Jay Sau, a physicist at the University of Maryland, describing what a logical qubit was supposed to be.
Here is what it has become, in the same breath: "Now, a 'logical qubit' seems to be a collection of qubits that is somewhat better than one qubit, and I'm not sure that the use for such logical qubits is as obvious."
A 'logical qubit' now seems to be a collection of qubits that is somewhat better than one qubit, and the use for such logical qubits is not obvious.
Hold that redefinition in your head, because it is the whole story.
On 22 June, President Trump signed two executive orders on quantum computing, surrounded by various tech leaders, including the president of Google and the CEO of IBM. The Department of Energy responded by committing to deploy "the world's first fault-tolerant, scientifically relevant quantum computing capability" by 2028.
Five years. World's first. Fault-tolerant. It sounds like a moonshot—a bet on hardware the government could plausibly lose.
Read what DOE actually targets and the deadline shrinks in front of you.
The goal, per DOE, is a machine with logical qubits "numbering in the low hundreds." That is the entire bar.
The current record holders are reportedly already close. QuEra reportedly claims 96 logical qubits. Quantinuum reportedly claims 94.
And "low hundreds" is not some distant aspiration—it is already a 2028 milestone on the published roadmaps of QuEra, IonQ, and IBM. The government has set a deadline that industry had already circled on its own calendars.
So the word "fault-tolerant" is doing enormous work while meaning almost nothing. Fault tolerance was supposed to mean a qubit that lives essentially forever, all errors corrected. What DOE is buying is a logical qubit that is "somewhat better than one qubit," in Sau's phrase—still noisy, just less so.
Edward Parker of the Rand Corporation puts the same point without the edge. "I think that achieving hundreds of logical qubits by 2028 is plausible," he says. "These will not be perfect logical qubits that are capable of sustaining very long computations."
How much better than a physical qubit? "How much more stable these logical qubits would be remains a very open question," Parker says.
That is the trap. If "fault-tolerant" now means "logical qubits somewhat better than physical ones, at a count already on commercial roadmaps," then the 2028 deadline will almost certainly be "met." It arguably appears designed to be met.
Which shifts all the weight onto two words nobody has defined: "scientifically relevant."
Here the experts split cleanly. Sau: "I cannot think of a scientific question that a quantum computer can help with in 2028, unless there is a serious breakthrough in device quality."
Pranav Gokhale, CTO of the quantum company Infleqtion, disagrees flatly. "100 logical qubits is where we see the advantage," he says, naming two targets: the model underlying magnetism, and the one that may underlie high-temperature superconductivity.
Notice that both men are talking about the same hardware. They are not disagreeing about qubit counts. They are disagreeing about whether a few hundred noisy logical qubits can answer a question worth asking—a definitional fight dressed up as a technical one.
And DOE has not settled it. The executive order gives the Secretary of Energy 90 days—until roughly late September—to identify and publicly release the technical specifications for what a "scientifically relevant" machine actually is.
Sit with that timeline. The government arguably promised the world's first scientifically relevant quantum computer before releasing, as appropriate, a technical specification for "scientifically relevant," which is due after the announcement.
This is why the qubit deadline is not the real news. It is a definitional maneuver with a near-guaranteed win baked in, and a victory lap already scheduled for 2028.
The real news is the structure of the vehicle—and it is unfunded.
DOE's Quantum Genesis initiative, a piece of the larger Genesis Mission, has three legs. A "DOE Q Competition" to demonstrate fault-tolerant systems in 2028. A National Quantum Supercomputing User Facility. And focused R&D to define the "keystone" scientific applications the machines will be judged against.
The order also tells Commerce to develop a plan "potentially including advance market commitments"—which, per FYI (aip.org), could involve government committing to buy or subsidize the technology once it exists.
But the orders and the DOE announcement do not specify how any of it gets paid for. They lean on public-private partnerships and industry contributions to fill the gap.
Congress has noticed. In its FY2027 appropriations report, the House Appropriations Committee said it would not fund additional AI supercomputers at DOE until it "gains a better understanding on infrastructure, operations, and outyear funding implications."
Some Democratic lawmakers have gone further, warning that the Genesis Mission could effectively pull federal dollars away from existing DOE research programs.
So the qubit deadline sits on top of an unfunded mandate that appropriators are already refusing to backstop. The hardware target is safe precisely because it was set low. The money is the actual uncertainty, and, per FYI (aip.org), its funding is unspecified and relies on public-private partnerships while Congress has requested briefings.
Now the half of this story that almost no one is covering.
The second executive order, per Kratsios, reportedly accelerates the federal migration to post-quantum cryptography—encryption designed to survive attacks by future quantum computers—to a 2031 deadline. The standards themselves already exist, released by NIST in 2024.
This is the consequential order, and it is the dull one.
Consider the asymmetry. The 2028 qubit goal is a research prize measured in the low hundreds of logical qubits, judged by a spec that does not yet exist, celebrated regardless of whether it answers a single real scientific question.
The 2031 crypto migration is a grinding infrastructure project: swapping out the cryptographic foundations of federal "high-value assets" across every agency, on a real clock, against a real threat model.
One is a definitional game with a guaranteed headline. The other is genuine engineering with a genuine deadline and no photo op.
Sau cannot name a scientific question quantum will answer by 2028.
The qubit deadline will probably be "met" on a technicality in 2028.
The crypto migration is harder, duller, and the one that actually protects anything. It is getting the least attention.