How David Silver’s Ineffable Intelligence Is Betting Against LLMs with a $1.1B Seed Round and a Google Cloud Supercomputer

The exclusive Google Cloud partnership

Six weeks after the funding closed, Ineffable Intelligence announced it had chosen Google Cloud to power its “superlearner” mission after a rigorous evaluation of the infrastructure market . The deal, revealed at the Google Cloud Summit London on June 16, 2026, gives the startup access to one of the world's largest arrays of NVIDIA-powered GPUs, specifically deploying on Google Cloud’s AI Hypercomputer architecture .

This partnership is a deep strategic alignment: Google was already an investor in the seed round, and the deal leverages Google Cloud’s latest AI-optimized hardware, including its eighth-generation TPUs and new A5X bare-metal instances powered by NVIDIA Vera Rubin chips, which were announced at Google Cloud Next '26 . The London summit location also underscores the UK's growing prominence as a hub for frontier AI research .

The “anti-LLM” thesis: reinforcement learning without human data

The central philosophy of Ineffable Intelligence is a direct challenge to the dominant large language model (LLM) paradigm. David Silver, a UCL professor and the former head of DeepMind’s reinforcement learning team, is the principal architect behind AlphaGo, AlphaZero, and MuZero—systems that mastered Go, chess, and other games through pure trial-and-error self-play, without needing to study human game records .

Silver’s core thesis is that LLMs, which are trained on vast corpora of human-generated text, have a fundamental ceiling: they can only interpolate from what humans have already written down. His proposed “superlearner” is an AI that learns entirely from its own experience, generating its own data through interaction with environments and discovering genuinely new knowledge that no human has ever documented . The company’s stated mission is “to make first contact with superintelligence” .

This approach eschews the pretraining on scraped internet text that underpins models like ChatGPT, Gemini, and Claude. Instead, it leans fully into reinforcement learning, a method where AI agents learn through trial, error, and reward signals . Silver’s 2024 paper “Reward is Enough” sketched out the intellectual framework for this bet, arguing that a sufficiently intelligent agent, driven by a reward signal, can derive all necessary knowledge .

A high-stakes bet on an unproven path

The immense funding and infrastructure commitment represent a landmark moment in the AI industry. The involvement of Nvidia, the primary GPU supplier, and Google Cloud, the infrastructure provider, alongside top-tier VCs, signals a significant institutional bet that reinforcement learning-based “experience learning” is a credible and potentially superior path to advanced AI . It positions Ineffable Intelligence as a counterweight to the LLM-focused strategies of companies like OpenAI and Anthropic.

However, the venture is not without significant risk. Reinforcement learning at the scale required for a general “superlearner” is extraordinarily computationally intensive and remains unproven outside of game-like or simulated environments. While the $1.1 billion seed provides a long runway, the company has yet to demonstrate that its approach can yield breakthroughs in real-world scientific discovery or complex, open-ended domains.