Investors pour billions into new AI labs before products exist

In short: Investors are putting tens of billions of dollars into new AI labs even when many have not released products or made revenue.

What's going on

A group of new companies building large AI systems, sometimes called “neolabs”, is raising unusually large amounts of money. The Financial Times reports that in the past two quarters, these labs raised about $24 billion, based on figures from Radical Ventures.

Some of these companies are getting very high valuations, which is the price tag investors put on a company. For example, Emulate, founded by former Google DeepMind researchers, is reported to be on track for a valuation close to $4 billion. Safe Superintelligence, started in 2024 by OpenAI co-founder Ilya Sutskever, raised money last year at a $32 billion valuation, even though it has not publicly shown a product.

These labs are expensive to build. They need top research talent and “compute”, which is the chips and servers used to run and train AI models (like renting a huge fleet of powerful computers). Because those inputs are scarce and costly, the companies often raise money frequently. Venture firm Chapter One estimates the median time between fundraising rounds is 7.6 months.

The Financial Times also notes that headline numbers can be misleading. Some investors may invest quietly at one valuation, then join a later, more public round at a higher valuation. That can make the company’s valuation look bigger, even if some money came in cheaper.

What to watch

Many of these startups are aiming at specific uses, like “world models” that simulate environments, or “physical AI” for machines and robots. A key question is which, if any, can turn large funding into real products that people and businesses pay for, and whether the lofty valuations hold up.

Source: Financial Times