Details
- Recursive, an AI company building recursive self-improving superintelligence, signed a multi-year, $410 million agreement to run its automated AI research system on Amazon Web Services as its primary cloud provider.
- The collaboration involves Recursive and AWS co-developing infrastructure purpose-built for large-scale, automated AI research, enabling parallel "research loops" that aim to compress months of sequential experimentation into days.
- Recursive’s system conducts open-ended experiments to safely improve itself and has already set new state-of-the-art results across three benchmarks (NanoChat, NanoGPT Speedrun, SOL-ExecBench), beating a two-year human leaderboard record and a live human-AI collaboration.
- The company emerged from stealth in May 2026 at a $4.65 billion valuation, backed by GV, Greycroft, AMD Ventures and NVIDIA, and was founded by former research leaders from OpenAI, Google DeepMind, Meta AI, Salesforce AI and Uber AI.
- According to external reports, Recursive raised about $650 million at launch and is routing a large share of capital into compute, reflecting a wider trend of AI labs prioritizing massive cloud contracts—with Amazon joining Microsoft and Google as key infrastructure partners for frontier AI research.
Impact
This deal underscores how self-improving and agentic AI approaches are driving a new wave of long-term, high-value cloud compute commitments, reinforcing hyperscaler dominance in frontier research infrastructure. Over the next 12–24 months, co-developed AWS tooling for automated AI research could influence how other labs structure their stacks, accelerate agent-based experimentation workflows, and intensify competition around scalable "AI scientists" and closed-loop optimization systems.
