When SpaceX absorbed xAI in February 2026, something quietly broke. xAI had been built as a research-first organization, pulling top scientists from DeepMind, OpenAI, and Google. SpaceX runs on a completely different rhythm: fast, hardware-driven, with Musk's signature pressure to ship. That culture clash pushed out virtually everyone who had built the place. All eleven original co-founders are now gone. More than 80 researchers and engineers left in the months that followed.
Igor Babuschkin was one of the first to go. He had been xAI's chief engineer, responsible for building the core systems that let the company train large AI models at speed. That kind of infrastructure work is unglamorous but foundational. You cannot build competitive AI without it, which is exactly what makes Babuschkin a credible name to investors.
He incorporated River AI in Nevada on April 20. Within weeks, he was in talks to raise up to $1 billion, with General Catalyst in discussions to lead the round. The target valuation is up to $5 billion. He is committing up to $100 million of his own capital. That personal stake matters: it tells investors this is not a trial run.
What makes this moment unusual is that River AI is not the exception. It is part of a pattern. David Silver, the researcher behind AlphaGo and AlphaZero at DeepMind, raised $1.1 billion in April for his startup Ineffable Intelligence at a $5.1 billion valuation. The company had no product, no revenue, and no public roadmap. Sequoia partners flew to London personally to secure the deal. Richard Socher, former chief scientist at Salesforce, is separately raising $1 billion for Recursive Intelligence. These ventures have a name now in investor circles: neolabs. They operate more like private research institutions than startups, with the freedom to work on ideas that would move too slowly inside a large company.
The money going into these labs is not small. In Q1 2026 alone, foundational AI startups raised $178 billion globally, double the total for all of 2025. That concentration of capital is creating a specific kind of pressure on every other industry: the best research talent is being pulled away from corporate labs and into these independent ventures, driving up the cost and scarcity of that expertise across the board.
For businesses that rely on AI to stay competitive, whether that is an insurance firm using AI to price risk or a procurement team using it to manage supplier contracts, this matters for a straightforward reason. The next wave of AI capabilities will be shaped by whoever wins this talent race. Neolabs are explicitly trying to build things that the large labs cannot, because they are too slow, too product-focused, or too constrained by commercial pressures. If any of them succeed, the gap between what AI can do and what businesses are currently using it for will widen again.
River AI has disclosed nothing about its research direction. That is normal at this stage. But Babuschkin's background points toward infrastructure and large-scale model training rather than consumer products. His career ran through DeepMind, OpenAI, and then xAI, always in the engine room rather than the front end.
General Catalyst is not a passive investor. It has been building AI-native companies and embedding them inside traditional service businesses across legal, healthcare, and business travel. Backing a foundational research lab would be a different kind of bet for them, one that suggests they believe the next layer of AI capability is still being written, not just deployed.
The real question River AI raises is not whether Babuschkin can raise the money. He almost certainly can. It is what happens in two or three years when these neolabs either produce something genuinely new or quietly wind down. That outcome will determine whether this funding wave was the beginning of the next AI era, or the most expensive research experiment in history.