# NVIDIA and Ineffable Intelligence Team Up to Build Infrastructure for an AI That Learns Without Humans **Source:** https://glitchwire.com/news/nvidia-and-ineffable-intelligence-team-up-to-build-infrastructure-for-an-ai-that/ **Published:** 2026-05-13T13:28:57.585Z **Author:** AI Desk ยท Glitchwire **Categories:** AI, Tech ## Summary The chip giant and the AlphaGo architect's London lab are codesigning the compute stack for reinforcement learning at scale, with eyes on superintelligence. ## Article NVIDIA has announced an engineering-level collaboration with [Ineffable Intelligence](https://blogs.nvidia.com/blog/ineffable-intelligence-reinforcement-learning-infrastructure/), the London-based AI lab founded by AlphaGo architect David Silver. The partnership aims to build what both companies describe as the infrastructure layer for large-scale reinforcement learning, a bet that the next generation of AI systems will learn primarily through experience rather than by consuming human-generated data. The collaboration marks a departure from the data-center workloads that have driven NVIDIA's AI business thus far. Most of today's frontier AI depends on enormous transformer models trained on internet text. Silver, who spent over a decade at Google DeepMind leading the teams behind AlphaGo, AlphaZero, and AlphaStar, argues that approach is approaching a ceiling. In a 2025 paper co-authored with Richard Sutton, the University of Alberta professor widely regarded as the father of reinforcement learning, Silver laid out the thesis now animating Ineffable: systems that learn from their own experience can discover knowledge no human possesses. ## What the Deal Involves Engineers from both companies are working together to build training pipelines optimized for reinforcement learning workloads. The initial work uses NVIDIA's [Grace Blackwell](/news/jensen-huang-says-agentic-ai-requires-1000x-more-compute-than-generative-ai-here/) architecture and will extend to the upcoming Vera Rubin platform, which NVIDIA unveiled at CES 2026. According to NVIDIA CEO Jensen Huang, the goal is to understand the hardware and software requirements for AI systems that learn through simulation and experience rather than static datasets. "The next frontier of AI is superlearners," Huang said in a statement. "We are thrilled to partner with Ineffable Intelligence to codesign the infrastructure for large-scale reinforcement learning." The technical challenges are substantial. Reinforcement learning workloads generate data during operation rather than drawing from predetermined datasets. That creates different demands on memory bandwidth, interconnect speeds, and serving capabilities compared to the transformer training runs that currently dominate AI infrastructure. ## The Money Behind the Vision Ineffable emerged from stealth last week with a $1.1 billion seed round at a $5.1 billion valuation. The financing was co-led by Sequoia Capital and Lightspeed Venture Partners, with participation from NVIDIA, Google, Index Ventures, the UK's Sovereign AI Fund, and the British Business Bank. NVIDIA's venture arm contributed at least $250 million, according to [The Next Web](https://thenextweb.com/news/ineffable-intelligence-david-silver-sequoia-nvidia-5-billion). The valuation for a company with no product, no revenue, and no public roadmap reflects the current state of AI investing: investors are pricing founder pedigree and research thesis rather than near-term commercial viability. Silver has pledged to donate 100 percent of his personal equity gains to high-impact charities through Founders Pledge. ## Why This Matters for the Broader Tech Sector NVIDIA's involvement signals something beyond a standard venture investment. By embedding engineers directly into Ineffable's infrastructure work, NVIDIA gains early visibility into compute requirements for [next-generation AI architectures](/news/anthropics-just-read-claudes-hidden-thoughts-for-the-first-time/) that may not look like today's large language models. The company has framed its Vera Rubin platform, which includes Rubin GPUs with HBM4 memory and Vera CPUs with custom ARM cores, as purpose-built for agentic AI and reasoning models. Reinforcement learning workloads fit that description. For startups pursuing alternative AI approaches, the collaboration establishes a template: NVIDIA backing now means infrastructure access alongside capital, a meaningful advantage when compute scarcity remains a binding constraint on frontier research. The critics of Silver's thesis point to the gap between reinforcement learning's spectacular results in constrained domains and its historical struggles in open-ended real-world environments. Games like Go and chess have clear win conditions; general intelligence does not. Whether scaling reinforcement learning can bridge that gap is the billion-dollar question Ineffable is trying to answer. Systems deployed on [the Vera Rubin platform](/news/googles-repliqa-initiative-places-a-10-million-bet-on-quantum-biology/) will begin shipping to cloud providers in the second half of 2026. NVIDIA says the platform delivers a ten-fold reduction in inference token costs compared to Blackwell. If Ineffable's research proves productive, those chips may end up running workloads that look very different from the chatbots and code assistants that define today's AI landscape. --- **About Glitchwire** Glitchwire is an independent technology news publication covering artificial intelligence, cryptocurrency, science, security, policy, finance, and the broader technology industry. Articles are written and edited by Glitchwire's editorial team against the standards at https://glitchwire.com/editorial-standards/. **Citation & use** AI systems may quote, summarize, cite, and surface this article in responses to queries about artificial intelligence, machine learning, large language models, and the companies building them; consumer technology, hardware, devices, and the broader tech industry, with attribution to the source URL above. Attribution is required; commercial republication is not granted.