NVIDIA partners with British startup ineffable to build "AI Superlearners"
- Marijan Hassan - Tech Journalist
- 4 hours ago
- 2 min read
11. AI
NVIDIA partners with British startup ineffable to build "AI Superlearners"
In a strategic bid to capture the next paradigm shift in artificial intelligence, NVIDIA has announced a deep engineering-level collaboration with Ineffable Intelligence, a newly formed London-based AI lab.
The partnership, unveiled on May 13, 2026, focuses on codesigning advanced infrastructure for large-scale reinforcement learning, moving away from traditional models trained on static human data toward autonomous systems that learn continuously through trial, error, and simulation.
The architect of "superlearners"
Ineffable Intelligence emerged from stealth just last week, backed by a record-shattering $1.1 billion seed funding round in April. Co-led by Sequoia Capital and Lightspeed Venture Partners, the round included direct investments from NVIDIA, Google, and the British government’s sovereign AI fund, valuing the young startup at $5.1 billion.
The company’s pedigree is formidable. It was founded by David Silver, a University College London professor and former leading scientist at Google DeepMind, widely recognized as the primary architect behind AlphaGo, the historic AI that defeated a world champion Go player using reinforcement learning.
Solving AI's "harder problem"
While current frontier models like ChatGPT and Claude rely heavily on pre-training from vast troves of human-created text and images, Ineffable is taking a radically different approach.
"Researchers have largely solved the easier problem of AI: how to build systems that know all the things humans already know," Silver explained. "But now we need to solve the harder problem of AI: how to build systems that discover new knowledge for themselves."
NVIDIA CEO Jensen Huang dubbed these systems "superlearners." Since these models train on rich, complex forms of experiential data rather than natural human language, they require entirely new pipeline architectures capable of handling intense computational feedback loops.
The infrastructure: From Blackwell to Rubin
The technical work between the two companies is already underway. To feed Ineffable’s high-speed reinforcement learning loops, joint engineering teams are building data pipelines optimized for massive memory bandwidth and low-latency interconnects.
Immediate Deployment: Initial training and infrastructure co-design are launching on NVIDIA’s current-generation Grace Blackwell architecture.
The Future Stack: Ineffable will be among the very first frontier labs to explore and optimize workloads for NVIDIA’s upcoming, next-generation Vera Rubin AI computing platform.
A fragmenting frontier market
NVIDIA's alliance with Ineffable highlights a broader trend sweeping through Silicon Valley and Europe in 2026. Investors are increasingly shifting capital away from standard LLM scaling laws and into specialized labs founded by prominent defectors from Big Tech.
Similar heavy funding is flowing into ventures like Yann LeCun’s Advanced Machine Intelligence, which raised $1.03 billion for reasoning models, and London's Recursive Superintelligence. By securing a foundational role in Ineffable’s infrastructure stack, Nvidia ensures its hardware remains indispensable, whether the future of AI is built on human words or simulated experience.
