Room Temperature at a Million Qubits: What Would Actually Have to Break First

The million-qubit line is not a marketing round number. It comes from a specific published result. In 2025 Craig Gidney showed that a 2048-bit RSA integer could be factored in under a week by a quantum computer with fewer than a million noisy qubits — a roughly twentyfold reduction from the 2019 Gidney-Ekera estimate of 20 million qubits for an eight-hour run.

The assumptions are worth knowing, because vendors quote the headline and skip them: a uniform 0.1% gate error rate, a one-microsecond surface-code cycle, ten-microsecond control reaction time, and a square grid of nearest-neighbour connections. The reduction came from approximate residue arithmetic, yoked surface codes for idle storage, and cheaper magic-state production — algorithmic and error-correction work, not hardware.

That is the honest shape of the field right now. The qubit requirement for the single most consequential quantum application has been falling because theorists keep finding better ways to spend the qubits, not because anyone built a bigger machine. An operator should expect that number to keep moving, and should not plan around any one estimate as a fixed date.

When a vendor says room temperature, ask which part. A quantum computer has qubits, control electronics, and readout. Those three can each be warm or cold independently, and the marketing rarely distinguishes them.

Nitrogen-vacancy centres in diamond are the modality where the qubits themselves are genuinely ambient — the carbon lattice insulates the spin state from thermal noise. SAXON Q has commercialised this, announcing a 128-qubit SXQ128 and a 512-qubit SXQ512 built as multi-core machines: eight fully entangled qubits per core on the smaller system, sixteen on the larger, coordinated across cores. The company claims fidelity up to 99.92% and says the systems plug into a standard electrical outlet with no cryogenic plant and no recalibration cycles. SXQ512 deliveries are slated for Q2 2027, with a stated roadmap toward 10,000-plus qubits.

Photonic approaches are the other ambient family — photons are naturally robust to thermal noise — but many photonic systems still need cryogenically cooled single-photon detectors. The qubits are warm; the machine is not. That distinction is the single most useful question to ask a quantum vendor.

Set 512 against a million and the gap is roughly three and a half orders of magnitude. That is not a scaling problem you solve with a bigger budget.

The dominant assumption is that cryogenics is the obstacle. It is an obstacle. It is not the wall.

The wall is fan-out. Superconducting and silicon spin architectures need control and readout lines per qubit. At a few hundred qubits you can run coaxial lines into a dilution refrigerator. At a million you cannot — not thermally, not physically, not economically. This is why the serious scaling work has moved to putting control silicon next to the qubits: milli-kelvin cryo-CMOS control chips demonstrated in Nature in 2025, work from TU Delft and Bluefors on cold control electronics, and imec's programme upscaling silicon spin qubits on 300mm wafers.

Now run the room-temperature scenario. You delete the refrigerator, the helium-3 supply chain, and the thermal budget on every control line. You do not delete the need to individually address, calibrate and read a million quantum elements. That remains an interconnect, packaging and test problem — and a warm machine arguably makes it harder, because you lose the thermal isolation that currently suppresses a whole class of crosstalk.

Anyone claiming a path to a million ambient qubits is making a claim about packaging and photonic or electrical I/O density. Judge it there, not on the qubit count.

If the scenario landed, quantum computing would stop being a cryogenics business and become a semiconductor business overnight. That is a bigger shift than it sounds.

Today the constraint is specialist: dilution refrigerators, microwave engineering, a small pool of people who can keep a machine cold and stable. In the warm scenario the constraints become the ones the industry already fights over — lithography capacity, wafer yield at scale, advanced packaging, photonic integration, and per-die test throughput. Quantum would be competing for the same fab and packaging capacity as AI accelerators, in the same constrained supply chain, against buyers with deeper pockets and shipping revenue.

For anyone with exposure to semiconductor supply — and by 2026 that is most manufacturers, aerospace suppliers and telecom operators, not just chip firms — that is the practical read. The quantum breakthrough scenario is a demand shock on wafer and packaging capacity, arriving in a market that has not had spare capacity in years.

Four questions separate a real result from a press release. Which components are at room temperature — qubits only, or the full stack including readout? Are the qubits physical or logical, and if logical, what is the physical-to-logical ratio? What is the two-qubit gate fidelity and the connectivity, not just the count — 512 qubits in cores of 16 is a different machine from 512 fully connected? And is error correction demonstrated or roadmapped?

Apply that filter and most of the quantum news cycle resolves into steady, real, unglamorous progress. Which is what it is.

There is no operational decision here for an SME. You should not be buying quantum hardware, hiring quantum engineers, or building a quantum strategy deck. There is exactly one thread from this scenario that carries a genuine deadline, and it is not compute — it is cryptography, timing and the infrastructure that carries them. That is the subject of the companion piece.