Google and Meta Are Losing Top AI Researchers as New Startups Rapidly Raise Billions

Big AI companies aren't just losing employees. They're losing well-known researchers with access to cutting-edge work. A growing number of these specialists are leaving Google, Meta, OpenAI and Anthropic to launch their own startups, and investors are ready to fund them almost immediately, often to the tune of hundreds of millions of dollars.

Why they're leaving

The generative AI boom has sharply raised the stakes for big labs. Companies with multibillion-dollar valuations have to ship new models regularly, push benchmark scores higher and turn research into products fast. That leaves strong scientists and engineers with less room for long cycles of fundamental work.

Investors and founders alike increasingly say that research freedom inside corporations is shrinking: releases, commercial KPIs and the race for leadership take priority over risky but potentially breakthrough directions. This is especially visible outside LLMs. When an entire organization is built around shipping the next model, applied and narrower research topics easily get pushed aside.

HV Capital partner Alexandre Joel-Carbonell puts it bluntly: performance pressure and fast release cycles leave almost no time for real research. For many, that becomes the turning point. If building a new lab is risky either way, it's better to do it at your own company than inside a giant.

Who has already raised money

New examples now appear almost every week, and this is no longer a story about small angel checks. In scale, rounds for new labs look more like financing for mature tech companies than seed capital for teams of a few people.

Most telling is that these companies are only a few months old. The market isn't buying revenue or product maturity in advance; it's buying the founders' reputations and the odds that they'll be the ones to build the next breakthrough stack.

• David Silver, a former Google DeepMind researcher, announced a $1.1 billion seed round for Ineffable Intelligence, a company that is only a few months old.
• Tim Rocktäschel, another DeepMind alum, is raising up to $1 billion for Recursive Superintelligence.
• AMI Labs, founded by a Meta alum, received $1 billion in March 2026, a few months after launch.
• Ricursive Intelligence, founded by former Anthropic and Google DeepMind employees Anna Goldie and Azalia Mirhoseini, raised $335 million across two rounds just a few months after its founding.

And these are no longer exceptions. Over the past year, former employees of OpenAI, DeepMind, Anthropic and xAI have raised large checks for new companies such as Periodic Labs, Ricursive Intelligence and Humans&.

According to Dealroom, venture investors have already put $18.8 billion into AI startups founded since the start of 2025 in 2026 alone. If the pace holds, the market will surpass the $27.9 billion raised by similar companies founded since the start of 2024.

Why investors are rushing in

The funds' logic is simple: in AI, value increasingly sits not in the legal entity itself but in a specific team and its ability to quickly build new models, products or research platforms. If a startup is launched by someone who was already behind notable breakthroughs at DeepMind, OpenAI or Meta, the market treats that as a near-guaranteed signal of quality.

Investors are afraid of missing the next big project, so they get in at the very earliest stage and are willing to pay for reputation, speed and access to rare talent.

There's a business factor too. A new independent company can sell technology to large customers without looking like an extension of one of the tech giants. For some markets this is critical, especially where sensitive data, chips or internal intellectual property are involved.

Ricursive Intelligence co-founder Anna Goldie put it starkly:

"For chipmakers to trust us with their valuable intellectual property, we have to be like

Switzerland."

In that case, a startup's neutrality becomes a commercial advantage.

What it means

The AI market is entering a phase where the main battle is not only over models and compute, but also over people who can build a new research school around themselves. For Google, Meta and other giants, it's a signal: salaries and infrastructure access alone are no longer enough to retain talent. For the industry as a whole, it's a sign that the next wave of strong AI companies may emerge not inside corporations, but from the people walking out of them.

*Meta is designated an extremist organization and banned in Russia.