NVIDIA Expands France Startup Investments as AI Moves Into Industry
NVIDIA plans investments in four French startups as companies apply its models and computing systems across healthcare, research, robotics and business operations.
France’s AI ecosystem is moving from infrastructure commitments to deployments across healthcare, materials research, robotics and business operations. NVIDIA’s latest planned investments in four startups—Doctolib, Entalpic, Genesis AI and Pigment—reflect that shift, while a wider set of French projects shows how companies are putting models and accelerated computing to work.
The practical cases vary widely. Some use AI to support clinical workflows or business planning; others apply models to physical systems, product imagery, scientific research and infrastructure management. The source gives few comparative performance details for the four startups, so the investments signal areas of interest more clearly than they establish measurable results.
Read All NVIDIA in Europe Announcements
Four startups, four application areas
NVIDIA says it has previously invested in French technology companies Alice&Bob, Gradium and Scintil Photonics. It has committed nearly $10 billion to the technology ecosystem across Europe. Its newly planned investments in Doctolib, Entalpic, Genesis AI and Pigment extend that activity into healthcare, materials discovery, robotics and enterprise planning.
Doctolib develops healthcare technology, including agents intended to give professionals more time to deliver care. Its patient-facing digital health companion is designed to improve access and support more continuous, preventive care. The source doesn’t specify how the agents are deployed, what tasks they handle, or how their effects on clinical work are measured.
Paris-based Entalpic focuses on engineering advanced materials at the atomic scale. The company has particular expertise in surfaces and interfaces used in semiconductor manufacturing and next-generation electronics. That focus makes materials discovery its central application, though no discovery benchmarks or development timelines are provided.
Genesis AI is building general-purpose robots, combining foundation models with data collection systems, simulation infrastructure and robot hardware. That description points to a broad technical stack: model development alone is not enough to produce a robot that can act in physical settings. The source does not detail the robots’ current capabilities or the conditions in which they operate.
Pigment, also based in Paris, sells business planning software. Its platform supports forecasting and scenario modeling across finance, sales, workforce planning and operations. These functions connect AI to decisions that span multiple business teams, but the source does not state whether the platform’s forecasts or planning workflows use specific models.
Compute and models for local deployments
France’s ecosystem has been built alongside compute commitments tied to the AI Action Summit in February 2025 and the Choose France Summits in 2025 and 2026. A related manufacturing plan comes from Bull, which is extending European AI server production to NVIDIA Vera Rubin Ultra NVL72 systems planned for 2028. Trays are to be produced in Europe, with rack integration at Bull’s Angers factory.
Another deployment takes computing beyond data centers. Altair-1, the first satellite in a planned 50-satellite constellation described as a $1 billion project, has reached orbit. Loft Orbital integrated the spacecraft; Orbitworks built it in Abu Dhabi. Orbitworks is a joint venture between Loft Orbital and Marlan Space. Altair-1 carries NVIDIA Jetson AGX Orin edge modules and runs Mistral models for Earth observation and autonomous operations in orbit. This setup brings processing onto the spacecraft, where the system can analyze data and act without relying on a continuous connection to ground infrastructure. The source provides no figures for onboard performance or the amount of data processed.
French organizations are also using NVIDIA Nemotron open models in deployments where they want to retain control over their data. LINAGORA and Docaposte secured the CANUT framework agreement, making Twake.ai, built on Nemotron, the referenced sovereign digital workplace for more than 2,500 French organizations. In financial services, SBS is developing a platform with NVIDIA that lets banks run governed generative and agentic services within their own environments.
Dassault Systèmes is applying Nemotron models and accelerated computing to its Industrial AI architecture. The work supports training its Industry World Models and developing Virtual Companions for industrial applications, grounded in science, physics and engineering. Mistral, meanwhile, introduced its open-weight Large 4 model in private preview. The source describes it as suited to cyber defense, manufacturing, finance and multimodal tasks, but supplies no benchmark scores. Fondation Inria is starting a working group on open-source advanced modeling and tool adoption, with findings expected in the spring of the following year.
Deployments put the technology to work
Several examples focus on research and commerce. L’Oréal is adding NVIDIA ALCHEMI to research workflows so scientists can test formulation possibilities computationally and guide laboratory experiments. The company says combining predictions with hands-on work can make discovery up to 100 times faster than traditional methods; that figure describes the discovery process, not a reported improvement for a particular product.
Mirakl has processed more than 70 million products using fine-tuned and open models running on NVIDIA GPUs and inference software. The company says seller catalog onboarding now takes less than two hours. For visual commerce, Photoroom and Vestiaire Collective are demonstrating the business effects of visual modeling on marketplace performance, though the source does not quantify those effects. Omi uses approved 3D digital twins with models optimized on NVIDIA to create product imagery for marketing, reporting four-times-faster inference.
In Côte-d’Or, ThinkDeep and Akila built a digital-twin agent for managing nearly 5,800 kilometers of roads and buildings. It uses NVIDIA Cosmos open world models, Nemotron and Omniverse libraries. The project has cut total operating costs by 10%, with a target of 30% as deployment expands. Port Sud de France uses ThinkDeep’s DeepBrain platform, powered by Nemotron and Cosmos, for goods processing and port security.
Other deployments test AI in settings where reliability and integration matter. Paris-based Enchanted Tools is deploying social robots to support patients, including in pediatric cancer care. Bouygues Telecom and Prisme.ai launched AI Studio V2, using NVIDIA’s full-stack infrastructure to let employees build agents in a controlled environment. In quantum computing, Alice & Bob has integrated NVIDIA NVQLink with its superconducting cat qubits to explore more efficient GPU-assisted quantum error correction. Quandela is using NVQLink to connect a DGX Spark system with photonic qubits and investigate hybrid quantum-classical applications.
What to watch
Together, these projects show a broad set of applications, but they are at different stages and report different kinds of evidence. Some provide deployment counts or speed and cost figures; others describe goals, platform capabilities or technical integrations without independent comparisons. The next useful signal will be whether these systems keep delivering measurable results as they move from specific deployments into wider use.
NVIDIA CEO Jensen Huang’s GTC Berlin keynote is scheduled for Wednesday, Oct. 21, at 11 a.m. CEST.
