Silicon Valley AI talent is accelerating their "departure" from big companies to start their own businesses, with Meta and Google becoming top talent "training grounds."

Top AI researchers are leaving tech giants like Meta and Google at unprecedented speeds, founding startups and rapidly securing massive funding, entering a new period of accelerated talent flow in Silicon Valley AI.

On April 28, CNBC reported that former Google DeepMind researcher David Silver announced that his startup, Ineffable Intelligence, founded only months ago, had completed a record-setting $1.1 billion seed round.

Another former DeepMind employee, Tim Rocktäschel, is reportedly seeking up to $1 billion in funding for his new company, Recursive Superintelligence. Meanwhile, after leaving his role as Meta AI chief, Yann LeCun’s AMI Labs completed a $1 billion raise in March of this year.

The report notes that investor enthusiasm is fueling this wave of departures. According to Dealroom data, since 2026, venture capital has injected $18.8 billion into AI startups founded from the beginning of 2025 onward, set to surpass last year's total of $27.9 billion.

Analysts highlight that not only do top talents take away technical know-how, but also deep insight into industry blind spots—this is the core logic behind investor bets.

Staggering Funding: Hundreds of Millions Raised Within Months of Founding

The report notes that the funding obtained by these departing founders has far exceeded the typical boundaries of early-stage investment.

Ricursive Intelligence, founded by former Anthropic and Google DeepMind researchers Anna Goldie and Azalia Mirhoseini, focuses on AI tools for chip design.

The company was founded in September last year and quickly completed two rounds totaling $335 million in December and January. Periodic Labs, founded by former OpenAI and DeepMind employees and dedicated to developing autonomous labs, completed a $300 million raise in September—just months after its founding.

San Francisco-based Humans& was founded in October last year by former Anthropic and xAI staff, and closed a $480 million raise in January.

Eurazeo's Stern attributes the competitive edge of these founders to their unique internal perspective:

"They know what truly works at scale, and they clearly know which opportunities are being abandoned internally. The opportunities are there."

Technological Differentiation: Betting on Next-Generation Paradigms Beyond LLMs

Reports indicate these emerging companies are not simply copying big-company routes, but are showing marked differentiation in their technology directions.

Joël-Carbonell of HV Capital points out that more and more AI researchers are questioning whether expanding the current large language model approach is sufficient to break through the next AI capability bottleneck.

AMI Labs is focused on developing AI systems capable of learning from continuous real-world data. A spokesperson for the company says,

"AI has made significant progress in content generation, but still falls short in fundamental cognition, causal reasoning, and reliable behavior in real-world environments. As AI moves from screens into industry, robotics, healthcare, and other physical contexts, these limitations become increasingly critical."

Ineffable Intelligence is focusing on reinforcement learning—enabling AI models to learn from experience rather than rely on human-labeled data, a contrast to the current mainstream internet text training approach. According to a source who spoke to CNBC, this is also the technical direction adopted by Humans&.

Ricursive Intelligence's Goldie emphasizes the strategic value of independence:

"For chipmakers to trust us with their most core intellectual property, we must remain neutral, something impossible inside Google."

Notably, the report points out that these startups, having secured ample funding, are now reaching back into big companies, creating a secondary flow of talent.

Goldie of Ricursive Intelligence reveals that the company has reassembled the core team of AlphaChip, "which involves recruiting some of our old colleagues." The current team’s backgrounds span Google, Anthropic, Nvidia, Apple, and xAI.

This model is common among many new companies—founders, relying on personal reputation and abundant funds from investors, are able to attract top researchers from previous employers and other AI giants, reinforcing talent competition between startups and corporates.

Behind the Departure Wave: Big Companies' "Involution" Creates Startup Windows

Arms races among large AI labs are inadvertently creating opportunities for smaller, more agile firms.

Elise Stern, Managing Director at French VC Eurazeo and an investor in AMI Labs, says,

"When you're in a race, you narrow your focus excessively. This creates a vacuum—entire research fields such as new architectures, agents, explainability, vertical models are being deprioritized. Not because they're unimportant, but because they can't win the immediate race."

Alexander Joël-Carbonell, partner at HV Capital, similarly points out that as major AI labs face pressure to justify sky-high valuations, commercial targets are increasingly prioritized, severely reducing the exploration space for top researchers.

"Inside large foundational model labs, the pressure to deliver benchmark performance and maintain rapid release cycles leaves little room for truly exploratory research—especially directions outside the mainstream large language model paradigm."

Analysis suggests this structural contradiction is pushing top talent wishing to pursue cutting-edge but non-mainstream research directions to increasingly prefer striking out on their own.

Risk Disclosures and DisclaimerThe market has risks, investment requires caution. This article does not constitute personal investment advice, nor does it consider individual users’ specific investment goals, financial situation, or needs. Users should consider whether any opinions, viewpoints, or conclusions in this article fit their own circumstances. Investing accordingly is at your own risk.