IBM's Starling quantum plan has reached the power-company stage
IBM's next big quantum milestone is no longer just a slide in a roadmap. It is a half-million-square-foot building, and a town planning board in Poughkeepsie wants to know whether the power will hold up. According to Mid Hudson News, IBM has proposed a roughly 511,000-square-foot facility on its Poughkeepsie campus to assemble and manufacture its next-generation Starling quantum systems.
That does not make IBM's 2029 fault-tolerant quantum target any more proven than it was last year. It does show the company has moved the story from keynote language to local infrastructure. The same planning-board discussion that surfaced the facility also surfaced the awkward question underneath it: IBM told the board existing grid capacity should be enough through 2030, but offered no projection beyond that, and board members asked for independent verification from utility Central Hudson, Mid Hudson News reported.
That is the part worth paying attention to. Quantum companies talk all the time about logical qubits, error-correction codes, and future data centers. Fewer of them reach the stage where a local board is asking for proof that the wires, substations, and long-term power planning will keep up. In June 2025, IBM said in its newsroom announcement that its IBM Quantum Starling system would arrive by 2029 in a new quantum data center in Poughkeepsie. An IBM technical blog post from the same announcement cycle laid out the larger pitch for large-scale fault-tolerant quantum computing, the still-hypothetical class of machines that would use error correction to keep enough qubits stable long enough to do useful work. The new facility filing is not independent confirmation that Starling will hit that date. It is evidence that IBM is spending real industrial effort behind the claim.
IBM's technical pitch for Starling is still ambitious. The company says the system will run 100 million quantum operations using 200 logical qubits, where a logical qubit means a more reliable qubit built by combining many error-prone physical qubits with error correction, according to IBM's June 2025 announcement. IBM also says its qLDPC error-correction approach, a family of codes meant to reduce the hardware burden of keeping qubits stable, cuts required overhead by about 90 percent compared with other leading codes, the company said. Those are IBM's numbers, not an external audit.
The local filing adds some concrete scale to the roadmap. Mid Hudson News reported that the project would demolish two existing buildings totaling 161,000 square feet and replace them with the new facility, with a projected workforce of about 200 employees. That is not lab-demo language. It is a company trying to build manufacturing and assembly capacity around a machine it says customers will use.
It is also why the utility question matters more than the square footage. If fault-tolerant quantum computing ever becomes a real industry, the bottlenecks will not all be inside the cryostat, the super-cold hardware container where quantum processors run. Some of them will look boring: land use, electrical service, cooling, and the ability to convince local stakeholders that a research promise can survive contact with physical infrastructure. IBM has not cleared that bar yet. The town board is still asking for receipts.
That leaves this as a useful but narrow signal. Quantum Computing Report's writeup largely repackaged the local planning-board story and IBM's older roadmap rather than adding original reporting. The real update is not that IBM still wants Starling in 2029. Readers knew that from last year. The update is that IBM's quantum roadmap has reached the stage where the limiting reagent may be as mundane as power-company paperwork.
That is not nothing. In quantum, the field is crowded with timelines that never have to meet a bulldozer.