Nvidia’s CUDA became the software moat around its AI hardware. France’s CEA wants to stop quantum computing from following the same path.
The government-funded research organization is working with quantum computing company Alice & Bob to integrate the startup’s software stack into hybrid quantum and high-performance computing (HPC) workflows through Bull’s Qaptiva platform.
The project would give future users another way to coordinate workloads across classical supercomputers and quantum processors, alongside Nvidia’s CUDA-Q. But it remains a development and research program, with no completed integration or performance results yet.
Dividing workloads between classical and quantum processors requires software that can coordinate the different resources and move work between them. Alice & Bob will integrate its software with Qaptiva using CEA’s experience in coupling quantum computers with HPC infrastructure.
Developed by French state-owned supercomputing company Bull, Qaptiva is a hardware-agnostic quantum application platform. It is designed to let developers write an application once and execute it across different quantum backends, either on-premises or through the cloud.
Its components include Qaptiva Access, which can schedule quantum jobs, route them to available quantum processing units (QPUs), and manage their execution within hybrid infrastructure.
Alice & Bob is developing fault-tolerant quantum computing (FTQC) systems based on “cat qubits,” which are designed to suppress bit-flip errors at the hardware level. The company unveiled Helium, its first quantum system, in June.
“Our priority is to make Alice & Bob’s unique FTQC technology usable in environments where real computational work happens,” Chloé Poisbeau, chief operating officer at Alice & Bob, said in a statement.
Alongside the infrastructure integration, Alice & Bob and CEA will develop algorithms for many-body problems, which involve interactions among large numbers of particles. The partners are targeting workloads that could run on early FTQC systems around 2030, with possible applications in materials discovery, chemistry, and industrial simulation.
The partnership also has a software-competition dimension. CEA executives told Reuters that supporting multiple frameworks alongside Nvidia’s CUDA-Q could help quantum computing avoid the software bottleneck that has reinforced Nvidia’s position in AI infrastructure.
Nvidia CUDA-Q is an open-source development platform for applications spa.nning CPUs, GPUs, and quantum processing units (QPUs). The distinction is therefore not simply between open and proprietary software. It concerns whether one vendor’s framework becomes the default orchestration layer for heterogeneous quantum-HPC infrastructure.
Nvidia’s mature libraries, development tools, and developer base have made CUDA the lowest-risk option for many AI workloads. At the same time, applications built around its ecosystem can be difficult and costly to move to competing accelerators.
CEA has already installed quantum systems from French vendors Quandela and Pasqal alongside its classical supercomputing infrastructure and plans to add an Alice & Bob system in 2027, according to Reuters.