Top researchers are jumping ship from big tech companies like meta and google As investors bet big on the commercial potential of early-stage AI labs, there is a push to launch startups and raise huge funding rounds in the process.
Amid massive spending on AI, many of these startups have raised hundreds of millions of dollars within months of being founded.
Yesterday, former Google DeepMind researcher David Silver announced that his months-old startup Ineffable Intelligence has raised a record $1.1 billion in seed funding. Tim Rocktäschel, another former DeepMind employee, is reportedly raising up to $1 billion for his new startup, Recursive Superintelligence. Rockteschel did not respond to CNBC’s request for comment.
AMI Labs announced a $1 billion raise in March, months after founder Yann LeCun announced he was stepping down from his role as head of AI at Meta. We develop AI systems that can learn from continuous real-world data.
Last year, former staffers from OpenAI, DeepMind, Anthropic, and xAI also raised hundreds of millions of dollars from investors for months-old ventures, including AI labs Periodic Labs, Ricursive Intelligence, and Humans&.
Many of these companies are themselves hiring heavily from their founders’ former employers and other AI giants, as investors provide the funding needed to lure top researchers from Big Tech.
Elise Stern, managing director of French VC Eurazeo, which backs AMI Labs, told CNBC that competition for AI supremacy among the biggest AI labs is creating opportunities for smaller, more nimble companies.
“During a race, your focus becomes narrower,” she added. “That creates a vacuum. Entire research areas like new architectures, agents, interpretability, vertical models, etc. are being deprioritized. Not because they aren’t important, but because they don’t win the immediate competition.”
strikeout
Investors are racing to pour money into AI labs founded by top researchers who used to work at big tech companies.
As of 2026, VCs have poured $18.8 billion into AI startups founded since early 2025, according to Dealroom. This is on track to surpass the $27.9 billion earned last year by companies starting in early 2024 or later.
Eurazeo’s Stern said the founders, who have worked at Frontier Labs, have “unique” insights.
“They know what works at scale, and they know exactly what’s left inside,” she said. “There’s an opportunity there.”
Alexander Joel Carbonell, a partner at HV Capital, which also invests in AMI Labs, told CNBC that as major AI labs seek to justify their astronomical valuations, an increasing focus on commercial goals is limiting the freedom of top researchers.
“Within large basic laboratories, the pressure to achieve benchmark performance and maintain rapid release cycles limits the scope for truly exploratory research, especially outside of the dominant LLM paradigm,” he added.
find the gap
Founded in September and raised $335 million in two rounds in December and January, Ricursive Intelligence builds AI tools to aid chip design. Founders Anna Goldie and Azalia Mirhoseini both previously worked at Anthropic and Google DeepMind, where they contributed to the AlphaChip project, which aims to automate chip design.
Goldie told CNBC that potential customers are likely to view the new company as a neutral partner rather than a competitor.
“For chipmakers to trust us with their most valuable intellectual property, we have to be in Switzerland, which would not be possible if we were at Google,” she added.
The company also turned to former colleagues. “We reassembled the core AlphaChip team, which included hiring some old collaborators,” Goldie said. Other team members have previously worked at Google, Anthropic, Nvidia, apple and xAI, she added.
Periodic Labs, founded by former OpenAI and DeepMind staffers, raised $300 million in September, a few months after launching. The aim is to develop an autonomous research institute.
HV Capital’s Joël-Carbonell told CNBC that a growing number of AI researchers are questioning whether further expansion of current large-scale language model (LLM) approaches is enough to reach the next level of AI capabilities.
AMI Labs, founded by former MetaAI chief Lakun, said, “Now is the right time, as AI has made great strides in content generation but still struggles with grounding, causality, and reliable behavior in real-world settings.”
Yann LeCun, chief AI scientist at Meta, speaks at the Viva Tech conference in Paris on June 13, 2023.
Chesnot | Getty Images News | Getty Images
“These limitations will become increasingly important as AI moves beyond screens and into industrial, robotics, healthcare and other physical environments,” a company spokesperson told CNBC.
Ineffable Intelligence focuses on reinforcement learning. Reinforcement learning is where an AI model learns from experience rather than human data, compared to many major AI models that are trained on internet text. a source familiar with the company told CNBC.
It’s an approach also taken by San Francisco-based Humans&, which was founded in October by former Anthropic and xAI employees and raised $480 million in January.
Google, Meta, Anthropic, and OpenAI did not respond to requests for comment.
