After more than thirty years of work, we are still celebrating logical qubits that endure for microseconds and pretending this is the foothill before a mountain already half climbed. It is not. It is a laboratory curiosity being dressed up as industrial inevitability. If one wishes to extrapolate honestly rather than ceremonially, then the answer is brutal: a quantum computer with roughly a thousand logical qubits, fully integrated, operating reliably for a month, is not twenty years away, or even fifty in any serious sense. On the present trajectory, it is a century or more away, and quite possibly never.

The reason is elementary, though one must speak it slowly for the benefit of the professionally optimistic. A microsecond is not a small month. It is not an early month. It is not a month with promise. It is millions upon millions of times smaller than the thing being promised, and that is before one even begins to account for the fact that a useful machine would not merely sit there existing, but would have to compute, correct errors, route signals, survive interference, maintain coherence, and do so across not one logical qubit but a thousand of them. That means the problem is not additive. It is multiplicative. Every component must work, every layer must hold, every correction cycle must continue, and every failure mode must be suppressed, not in one toy device but across an integrated machine of monstrous complexity.

This is where the fraudulence of the public narrative becomes obvious. We are invited to imagine a smooth curve of progress from primitive demonstration to world-changing machine, as though the remaining distance were just more of the same. But it is not more of the same. It is an entirely different class of achievement. Keeping one fragile logical qubit alive in a controlled experiment for a trivial span is one thing. Building a system with a thousand logical qubits, which in practice implies vast physical overhead, with all the control electronics, cooling, fabrication precision, connectivity, synchronisation, and fault tolerance required to run continuously for a month, is something else altogether. It is the difference between making a paper aeroplane glide across a room and declaring commercial intercontinental aviation to be imminent.

So if thirty years has brought us here, then what does extrapolation actually say? It says the remaining task is not another neat thirty-year block, because the last thirty years did not solve the hard part. They merely established how far away the hard part still is. A charitable estimate gives a century. A more honest one gives several. A rational one admits that many technologies never cross the gulf between experimental novelty and practical civilisation-scale utility. They simply become grant-eating monuments to human overstatement.

That is why the correct estimate is not merely “a long time.” It is that there is no visible engineering path from microsecond-scale logical demonstrations to month-long, thousand-logical-qubit computation that fits within any normal strategic horizon. Not a decade. Not two. Not five. On present evidence, the sensible answer is a hundred years or more, if ever, because the gap is not one of enthusiasm, but of orders of magnitude, and arithmetic has a tiresome habit of refusing to be bullied by marketing.

1/3
S. Tominaga, AKA CSW
Apr. 6, 2026
https://x.com/i/status/2041004880089989155

https://t.me/S_Tominaga/5241