Mirendil closes $200M seed at $1B on bet that AI can automate AI research

Mirendil, a San Francisco-based frontier lab founded by former Anthropic researchers, raised $200 million in seed funding on June 24, 2026, at a $1 billion post-money valuation, one of the largest seed rounds in AI history. The round was co-led by Andreessen Horowitz and Kleiner Perkins, with participation from NVIDIA and others. Co-founders Behnam Neyshabur and Harsh Mehta left Anthropic in December 2025 after joining in late 2024. The team also includes Shayan Salehian (early xAI) and Tara Rezaei (MIT), plus about 20 researchers and engineers recruited from Anthropic, xAI, Google DeepMind, and OpenAI.

Mirendil's mission is to build AI systems that automate AI research itself—automating experimental design, hyperparameter search, model evaluation, and iterative training cycles. Rather than creating general-purpose foundation models, the company aims to develop specialized models and research workflows that let scientists, research institutes, hospitals, and enterprises build custom AI for their domains (drug discovery, chemistry, biology, robotics, materials science) without maintaining large ML engineering teams. The $200 million will fund GPU compute clusters, specialized scientific data repositories, and aggressive hiring of elite researchers.

The funding reflects a structural shift in venture capital: traditional metrics like revenue, customer growth, and near-term monetization are being set aside in favor of evaluating scientific credibility, technical ambition, and the quality of prior work at frontier labs. Mirendil has no product, no revenue, and no disclosed technical roadmap—yet attracted capital comparable to late-stage startups. NVIDIA's participation signals the company's strategy to invest across the emerging AI lab ecosystem: more frontier AI research means more demand for high-end GPUs used to train advanced models.

For research institutions and enterprises betting on AI, Mirendil's vision addresses a real gap: the frontier AI capabilities developed at a handful of mega-labs (OpenAI, Anthropic, Google DeepMind) remain out of reach for universities, pharmaceutical companies, and smaller research organizations. If Mirendil successfully builds tools to democratize frontier AI research, the market opportunity could be substantial—but the technical challenge (recursive self-improvement of AI systems) remains unproven at scale. Product launch expected in coming months.