AI startups focused on long-term research breakthroughs secured $24 billion in funding over the past two quarters, a capital surge that marks a sharp departure from the venture model that built OpenAI and Anthropic.
These so-called NeoLabs—companies founded by researchers departing OpenAI, Google DeepMind, Meta and top universities—operate without established products, revenue or near-term commercialization paths. They function as privately funded research institutions with the autonomy and capital to pursue ideas that incumbent firms avoid. Over 40 such labs have raised $40 billion in total over the past three years, with many commanding billion-dollar first-round valuations.
The $24 billion raised in two quarters alone exceeds the $5.3 billion that OpenAI and Anthropic collectively raised across seven years before ChatGPT's launch. The funding model assumes investors will hold positions for five to seven years before any commercial signal emerges.
Safe Superintelligence, founded by Ilya Sutskever in June 2024, exemplifies the trend. The company raised $1 billion within four months and reached a $32 billion valuation without a public product roadmap. Sutskever left OpenAI, where he served as chief scientist, to pursue what he described as "one goal and one product: a safe superintelligence."
Other researcher-led labs—Thinking Machines, World Labs and Reflection—emerged from departures of elite scientists seeking environments focused purely on technical exploration rather than commercial scaling. The migration reflects growing consensus among frontier researchers that foundational breakthroughs require institutional independence from product-driven constraints.
OpenAI established the prototype for this model. Founded in 2015, the company validated Ilya Sutskever's thesis that sufficiently large neural networks trained on massive datasets would yield capabilities beyond smaller, manually engineered systems. GPT-1, GPT-2 and GPT-3 proved the approach; ChatGPT's launch cemented it as industry orthodoxy.
Anthropics and Cohere followed distinct but instructive paths. Cohere, founded in 2019 by Aidan Gomez, Ivan Zhang and Nick Frosst, built enterprise-grade infrastructure for production language model deployment. Anthropic, formed in 2021 by Dario and Daniela Amodei and five researchers departing OpenAI, pursued scaling with explicit emphasis on safety, achieving state-of-the-art results.
The NeoLabs wave challenges the narrative that settled after ChatGPT: that capital requirements had become prohibitive for startups and that talent would concentrate within a handful of incumbents. Venture firms including Radical Ventures have raised dedicated $1 billion-plus vehicles to back these companies, signaling conviction that foundational breakthroughs—not near-term revenue—remain the capital allocation imperative in frontier AI.
