Updated on
In August 2025 IBM and Google targeted million-qubit machines by 2030. A year on, IBM's modular cooling step shows how much work remains.
In August 2025, IBM and Google told the Financial Times that they expect to deliver industrial-scale quantum computers by the end of the decade. Some headlines read that as a breakthrough. It was not: it was a stated goal, with no million-qubit machine behind it.
The gap matters. Today's experimental systems have fewer than 200 qubits, and IBM's Condor chip, at 433 qubits, ran into crosstalk interference when it was scaled up. As Slashdot summarised the report, IBM then moved to low-density parity-check error correction, which needs up to 90% fewer qubits than Google's surface code, but requires longer connections between distant qubits.
Google's stated plan is about cost rather than qubit count: cut component costs tenfold to keep a full machine near $1 billion. That is a procurement target as much as a physics one. And the sceptic in the story is not a bystander: Oskar Painter of AWS put useful quantum computers 15 to 30 years away, arguing the fundamental physics is largely solved but the engineering is not.
What happened next
A year later there is at least one concrete engineering data point. On August 19, 2026, IBM announced that it had joined and cooled two cryogenic modules into a single environment. Together they stand more than 8 feet tall and 8 feet wide, cooled to 4 kelvin in under five days and then below 15 millikelvin shortly after.
IBM says it will use L-couplers in 2027 to link processors into a system with at least 1,000 programmable qubits, and it still expects to deliver its fault-tolerant IBM Quantum Starling in 2029.
Notice what changed and what did not. The 2029 date is the same, and the next stated step is 1,000 qubits, not a million. Linking fridges is real progress on the longer-wire problem that IBM's error-correction approach creates. But it is a milestone on a roadmap, announced by the company that wrote the roadmap.
Our take: both camps can be right at once. The physics may well be settled while the plumbing, wiring and cost decide the schedule. We would treat 2030 as a vendor target and AWS's estimate as the counterweight, and judge the field by delivered modules rather than headlines.
Related articles from LaRebelión
Apple tightens full-disk access to rein in AI agents
OpenAI probes more possible AI hacking incidents after the Hugging Face breach
OpenAI Reports Rogue AI Agents Affecting Over 100 OrganizationsOriginal source: The Quantum Insider
Produced with AI support and reviewed by the newsroom