Every year we ask the same question: which AI companies actually matter?
This year, it got a lot harder to answer. Not because there were fewer contenders, but because the market is maturing across multiple fronts at once.
The CB Insights AI 100 identifies the most promising early-stage AI companies based on market traction, investor quality, talent signals, and our proprietary predictive insights.
And for the first time ever, the report is interactive, helping you find the companies that matter even quicker.
Here's what stood out this year:
1. AI agents need their own rulebook. Companies like Prophet Security and Bretton AI are running agents through over a million production workflows without human sign-off. The infrastructure for governing those agents is still getting built: identity, credentialing, and accountability for autonomous systems. Many of the companies featured are doing exactly that.
2. Physical AI enters the list as a standalone category for the first time, with 11 companies spanning robotics software, autonomous hardware, and enabling chips making the cut. Physical AI raised a record $78B in 2025. The next frontier is fleet coordination, with companies like FieldAI ($314M Series A, $2B valuation) and InOrbit.AI (200% customer growth last year) setting the foundation.
3. Vertical AI moats are about data, not sector. The companies pulling ahead sit on data that is difficult to replicate, whether through proprietary model training or embedding into workflows with real switching costs. Salient says it plainly with its 0% churn.
The 100 companies on this list are solving real problems with evidence of traction. In a market where AI valuations are heavily inflated and bubble concerns are real, that combination is rarer than one might think.
The full report is free to access. Link in the comments below.
Massive thanks to Adya Pandey for leading the effort, as well as the rest of the team that made it happen!
