Core Automation eyes $300M–$500M raise at $4B valuation — six weeks after launch
What's the deal? Core Automation, an AI model developer founded in late March 2026 by former OpenAI VP of research Jerry TworekDealroom has a profile for this one. Try Dealroom →, is seeking to raise between $300M and $500M at a valuation of around $4B, The Information reported on May 7, 2026. The discussions are early and terms could change. The raise comes just weeks after the company closed $100M at a $1B valuation — a rapid back-to-back fundraise that underscores soaring investor appetite for new AI labs.
Nvidia backed the earlier round, alongside Spark Capital and Accel. Tworek spent nearly seven years at OpenAI, where he led work on reinforcement learning and reasoning models, before departing in January 2026. He co-founded Core Automation with ex-Anthropic and Google DeepMind researcher Rohan Anil and Joanne Jang, who previously led OpenAI's work on model behaviour; other cofounders include Google DeepMind's Anmol Gulati and former OpenAI Chief People Officer Julia Villagra.
Why now? Nvidia's investment spree in startups that use its AI chips has turbocharged venture interest in those same companies. That dynamic is enabling brand-new labs — so-called "neolabs" — challenging OpenAI and AnthropicDealroom has a profile for this one. Try Dealroom → to stack funding rounds within weeks of each other, often before shipping a product. Core Automation isn't close to releasing one. It has also been recruiting top researchers from AnthropicDealroom has a profile for this one. Try Dealroom → and Google DeepMind.
What it's building. Core Automation wants to build the most automated AI lab in the world, starting by automating its own research. Rather than chasing ever-larger models trained on more data, it's pursuing continual learning — models that keep learning from real-world experience after training — plus architectures designed to outscale transformers. It's taking unorthodox approaches such as combining model-training stages into one and using far less training data than rivals.
The signal: The trajectory reflects a broader pattern — investors placing massive bets on small teams led by elite AI researchers, wagering that pedigree and novel approaches can challenge entrenched players. Going from $100M to a potential $500M in a matter of weeks suggests the AI-lab arms race is entering a phase where founding-team credentials alone can unlock billions in valuation before a product reaches market. Peers have shown the risk too: Thinking Machines Lab raised at $10B five months after founding but has since lost cofounders, and Safe Superintelligence lost its CEO to Meta.
Read more: The Information · The Decoder
