Leading AI researchers are increasingly leaving Big Tech to start their own labs. Investors are actively supporting such initiatives, pouring billions of dollars into newly formed teams focused on niche developments and new AI architectures.
Leading AI researchers are increasingly leaving Big Tech to start their own labs. Investors are actively supporting such initiatives, pouring billions of dollars into newly formed teams focused on niche developments and new AI architectures.
As CNBC reports , in the last year alone, Google-backed startups DeepMind, Meta, OpenAI, and Anthropic have raised record amounts of funding.
For example, former DeepMind researcher David Silver raised $1.1 billion in a seed round just a few months after launching his startup Ineffable Intelligence. Another former DeepMind employee, Tim Rockteshall, is reportedly raising about $1 billion for his Recursive Superintelligence project.
Experts say the reason for this situation is the excessive commercialization of large corporations. According to Elise Stern, a leading director at the venture fund Eurazeo, the race for dominance is forcing giants to narrow their focus.
“Whole areas of research, such as new architectures, agents, and model interpretability, are shifting priorities. Not because they are unimportant, but because they are not winning the race in the short term,” she noted.
The trend is also reinforced by the transition of Yann LeCun, who left his position as head of AI at Meta to develop AMI Labs . His new project, which has already raised $1 billion, focuses on systems that can learn from real data, not just text from the Internet.
Meanwhile, startup Ricursive Intelligence, founded by former Anthropic experts Anna Goldy and Azalia Mirhosseini, is developing tools for chip design. Goldy emphasizes that its status as an independent startup allows it to be “neutral territory” for customers who see Google as a competitor.
According to Dealroom, in 2026, venture capitalists have already invested $18.8 billion in AI startups created since the beginning of 2025. This indicates that the market is ready to fund alternative approaches that go beyond traditional large language models.
Previously, Yann LeCoun said that the US could lose technical talent due to Trump's policies.
