DARPA now tests six quantum companies' roadmaps against their hardware

A quantum sales meeting tends to end on a roadmap slide, with qubit counts rising year by year toward a box marked "utility scale". The person who signs off on the pilot budget has mostly had the vendor's word for whether that box can be reached.

In Stage C, a government team checks the slide. "Our team will be closely scrutinizing the hardware and systems as they develop to determine whether the technology performs the way the models and roadmaps say it should," said Micah Stoutimore, managing director of DARPA's Quantum Benchmarking Initiative ^[3]. Each design is tested on whether it can be built as planned and works as intended ^[5]. Stage B reviewed plans. Stoutimore described it as "putting the performers' plans under a microscope" ^[7]. Microsoft and PsiQuantum arrived through a pilot called US2QC, whose final phase has the same technical goals as Stage C, according to DARPA ^[8]. That makes six companies in Stage C ^[18].

Buyers building a shortlist will read the list as a ranking. DARPA says the programme does not pick winners ^[4], and describes its aim as trying to "spot every company on a plausible path to a useful quantum computer" ^[6]. Seven Stage B participants missed this round, and DARPA said that does not necessarily mean they won't advance ^[11]. Of the eleven Stage B participants named in DataCenterDynamics' report, four moved up ^[19]. The agency's forecast is about the field, too. "We increasingly expect that someone will build a utility-scale quantum computer by 2033," Stoutimore said ^[14].

The gap between what is pitched and what is checked shows in Atom Computing's own announcement. The company says it has built systems with more than 1,200 qubits ^[16] and has signed a $100m letter of intent with the US Department of Commerce ^[17]. It also says the stage carries potential funding of up to $300m ^[12]. "We believe utility-scale quantum computing is closer than many expect," said Ben Bloom, its chief executive and founder ^[13]. Microsoft's June claim that agentic AI made its quantum chip 1,000 times more reliable is likewise the company's own account ^[15].

DARPA counts a quantum computer as useful when its computing value exceeds its cost ^[10]. A 1,200-qubit system is a hardware fact ^[16]. Whether any machine clears that cost bar by 2033 is what the programme was set up to find out ^[1].

For the pitch on Monday, sort each claim on two axes: who checked it, the vendor or DARPA's team, and what it describes, a plan or running hardware. A vendor's own plan is the roadmap slide. A plan DARPA has reviewed, through Stage B or the US2QC equivalent, is a path a government team examined before moving it to hardware checks [c16, c17]. Vendor-reported hardware is where Atom's qubit count and Microsoft's reliability figure sit [c14, c13]. DARPA-checked hardware is the cell Stage C exists to fill ^[3].

In my view, Stage C status is a reasonable floor for anything past an exploratory contract. Spending on vendor-reported hardware is research money until DARPA's team says the hardware performs as the roadmap says. That rule costs most for buyers whose best-fit vendor is among the seven Stage B participants outside this round. DARPA says those companies may yet advance, and expects more to reach Stage C soon ^[11].