Overview
- Ineffable Intelligence announced a strategic deal with Google Cloud to be its preferred infrastructure partner and to deploy one of the largest NVIDIA Vera Rubin NVL72 (A5X) GPU clusters on Google’s AI Hypercomputer.
- The company says the setup uses Google Cloud’s integrated stack—Jupiter high-performance networking and optimized storage—because experience-based reinforcement learning needs tightly orchestrated networking, storage and GPU coordination rather than simple chip rental.
- Ineffable closed a $1.1 billion seed round in April that includes top venture investors and the UK Sovereign AI Fund, which underwrites the project’s London base and helps attract engineering talent to Europe.
- Ineffable is working with NVIDIA on engineering requirements to make the GPU environment scale for next-generation reinforcement-learning algorithms, but independent technical validation and details on safety and governance remain unresolved.
- If successful, the effort aims to build a ‘superlearner’ that learns from its own actions rather than human datasets, a path that could change how models discover knowledge but that is unproven and poses major technical and safety questions for regulators and researchers.