An 18 qubit Alice & Bob system, marketed as the first state acquisition of its kind, will be installed at a French supercomputing center in 2027. The contract value, access rules, and the vendor benchmark claims behind the purchase have not been disclosed.
France just signed a public check for a quantum computer. The deal, announced at VivaTech 2026, is being framed as the first formal state acquisition of an error-biased cat-qubit system anywhere in the world, according to the Quantum Computing Report coverage of the announcement. What the contract actually buys the country, and on what terms, is less clear.
GENCI, France's national high-performance computing agency, signed the procurement with Paris-based hardware developer Alice & Bob for an 18-cat-qubit machine. The deal is funded entirely through the France Hybrid HPC Quantum Initiative (HQI), part of the broader France 2030 investment plan, and it positions France as a customer of an architecture it did not develop.
The system will be physically installed at the Très Grand Centre de Calcul (TGCC), the CEA-operated computing facility in Bruyères-le-Châtel south of Paris. Researchers are expected to get access in 2027, initially through a hybrid setup with GENCI's Joliot-Curie supercomputer, with a later phase connecting the quantum machine to Alice Recoque, the European exascale system being acquired through the EuroHPC Joint Undertaking.
The machine itself is small. Eighteen cat-qubits is not a utility-scale system; it is a research instrument designed to test the architecture rather than run production workloads. Alice & Bob describes it as the first early Fault-Tolerant Quantum Computer (eFTQC) permanently installed at a European supercomputing center, a designation that frames the deployment as a stepping stone toward France's PROQCIMA program, which targets two 1,024-qubit universal quantum prototypes by 2032. The Quantum Computing Report write-up of the announcement carries the same framing.
The architectural pitch is that cat-qubits suppress bit-flip errors natively, which the vendor says reduces the physical-qubit scaling overhead by up to 200 times compared with conventional transmon systems. Bit-flip protection lifetimes exceeding an hour are also cited. Both numbers are vendor claims, not independently benchmarked results, and the trade-press framing of the buy as a "world's first" comes from the same source chain.
That attribution matters, because the public rationale for spending public money on a single-vendor architecture, with chips fabricated outside the buyer's territory, depends on the superlatives holding up. France's framing of the purchase as a piece of "European strategic autonomy" infrastructure is, in practice, a bet on Alice & Bob's roadmap and on the access policy that GENCI has not yet published.
The operational questions the announcement does not answer will determine whether the buy is judged a success or a press release. Which French and European researchers will get time on the machine, on what selection mechanism, and at what cost to their institutions? What benchmarks will be run against Joliot-Curie classical workloads, and who audits them? What does sovereignty mean when the architecture is single-vendor and the access policy is opaque?
Until GENCI publishes which researchers get time on the machine, on what selection mechanism, and at what cost, the public case for the buy rests on vendor numbers. The procurement price has not been disclosed. The next thing worth watching is the access policy, not the 2027 ribbon-cutting.