Nvidia has announced a partnership with AI startup Ineffable Intelligence, a company founded by former Google DeepMind reinforcement learning leader David Silver to pursue next-generation AI systems based on reinforcement learning.
The partnership will focus on building infrastructure for large-scale reinforcement learning systems that learn through experience and trial-and-error rather than relying primarily on human-generated data. Nvidia said engineers from both companies will collaborate directly on infrastructure and training pipelines optimized for reinforcement learning workloads.
The London-based startup was founded in late 2025 and emerged publicly earlier this year with a record $1.1 billion seed funding round co-led by Sequoia and Lightspeed. Investors included Nvidia, DST Global, Index Ventures, Google, and the U.K.’s Sovereign AI Fund.
Nvidia said the collaboration will use its Grace Blackwell chips alongside the company’s upcoming Vera Rubin platform to support large-scale training environments.
“The next frontier of AI is superlearners — systems that learn continuously from experience,” Nvidia CEO Jensen Huang said in a statement.
Silver said current AI systems have largely mastered learning from existing human knowledge but still struggle to independently discover new knowledge and strategies through experience.
“Researchers have largely solved the easier problem of AI: how to build systems that know all the things humans already know,” Silver said. “But now we need to solve the harder problem of AI: how to build systems that discover new knowledge for themselves.”
According to the company, Ineffable’s systems may require new model architectures and training methods beyond conventional large language model approaches.
The companies said they will specifically focus on creating scalable pipelines capable of continuously feeding reinforcement learning systems with simulated experiences and environmental feedback.
Reinforcement Learning Returns to Center Stage
Reinforcement learning previously played a central role in major AI breakthroughs including DeepMind’s AlphaGo systems, but recent generative AI advances have been driven largely by large language models trained on massive datasets of human-created text and images.
Several leading researchers now argue that future AI progress may increasingly depend on systems capable of independently exploring environments, testing strategies, and learning from outcomes rather than relying solely on static datasets.
That transition requires significantly different infrastructure, training pipelines, and computational architectures compared with conventional generative AI systems.
New AI Labs Intensify Competition For Talent And Capital
Ineffable is part of a broader wave of new AI research companies launched by former researchers from major AI labs including OpenAI, DeepMind, Anthropic, Meta, and xAI.
Investors have poured billions into these startups over the past year as competition intensifies around superintelligence research, autonomous AI systems, and advanced reasoning models. On the same day as Nvidia’s announcement, AI startup Recursive Superintelligence, founded by former DeepMind researcher Tim Rocktäschel, announced a $650 million funding round.
The surge in funding reflects growing investor belief that the next phase of AI competition may depend not only on larger models, but also on entirely new learning methods, infrastructure systems, and training paradigms capable of producing more autonomous forms of intelligence.
