NVIDIA taps British startup

NVIDIA said on May 13 that it had entered an engineering-level collaboration with London-based startup Ineffable Intelligence to build infrastructure for large-scale reinforcement learning, a form of AI training in which systems learn through trial and error rather than from fixed human-created datasets. The work starts on NVIDIA’s Grace Blackwell systems and will be among the first to explore the company’s upcoming Vera Rubin platform, according to an NVIDIA blog post. A separate filing disclosed on May 13 that the Jensen and Lori Huang Foundation is buying computing time from CoreWeave and donating it to universities and nonprofit institutes, with the gift valued at $108.3 million so far. The computing resources are earmarked for science and artificial intelligence research, and NVIDIA said in the filing it plans to offer free engineering services to some recipients. (blogs.nvidia.com) ### Which British startup did NVIDIA choose? Ineffable Intelligence is a London-based AI lab founded by David Silver, a former Google DeepMind researcher and a University College London professor focused on reinforcement learning. The company emerged from stealth on April 27 with a $1.1 billion seed financing led by Sequoia Capital and Lightspeed at a $5.1 billion post-money valuation, with participation from NVIDIA, Google and others. (money.usnews.com) CNBC reported that Silver founded the company in late 2025 and that Ineffable is pursuing systems that learn from experience rather than relying mainly on human data. Reuters reported on April 27 that the round was Europe’s largest seed financing to date. ### What are NVIDIA and Ineffable building together? NVIDIA said the two companies are co-designing a training pipeline for reinforcement learning systems at scale. (cnbc.com) The company said those workloads differ from pretraining because they generate data on the fly and require systems to act, observe, score and update continuously, putting pressure on interconnects, memory bandwidth and serving. Jensen Huang said, “The next frontier of AI is superlearners — systems that learn continuously from experience,” while Silver said researchers now need systems that “discover new knowledge for themselves.” NVIDIA said engineers from both companies are working together on the pipeline and that the effort will begin on Grace Blackwell before moving onto Vera Rubin exploration. ### Why do Grace Blackwell and Vera Rubin matter here? (blogs.nvidia.com) NVIDIA said on March 16 that Vera Rubin is a seven-chip platform intended to support every phase of AI, from pretraining and post-training to test-time scaling and agentic inference. The platform combines the Vera CPU, Rubin GPU, NVLink 6 Switch, ConnectX-9 SuperNIC, BlueField-4 DPU, Spectrum-6 Ethernet switch and a newly integrated Groq 3 LPU, according to the company. (blogs.nvidia.com) The May 13 collaboration gives Ineffable early access to NVIDIA’s current and next-generation AI systems for reinforcement-learning workloads. CNBC reported that the joint effort will use Grace Blackwell chips alongside Vera Rubin as the companies try to build AI systems that learn by trial and error. ### What does the $108 million donation actually fund? The May 13 filing said the Jensen and Lori Huang Foundation is purchasing cloud computing capacity from CoreWeave and donating it to universities and nonprofit institutes. (nvidianews.nvidia.com) Reuters reported that NVIDIA had previously invested $2 billion in CoreWeave in January and signed a $6.3 billion cloud-capacity deal with the company last year. (cnbc.com) NVIDIA said some grant recipients will also receive free engineering services. The filing did not, in the material reviewed, name the universities or nonprofit institutes receiving the donated compute. ### What comes next? NVIDIA’s newsroom says the company will hold a conference call on May 20 at 2 p.m. Pacific time to discuss first-quarter fiscal 2027 results. (money.usnews.com) Ineffable’s collaboration work is already starting on Grace Blackwell systems, with Vera Rubin identified by NVIDIA as the next platform to be explored in the project. (nvidianews.nvidia.com)