Y Combinator CEO Garry Tan is offering a vivid glimpse into the future of AI-powered work. In a new podcast appearance, Tan said he uses AI agents so heavily that he can spend up to $200,000 a year on tokens, underscoring just how central these tools have become for some of the tech industry’s most active builders.
While Tan joked about being so excited by AI agents that he sometimes barely sleeps, the bigger story is the level of value early adopters are already seeing. For startup leaders, engineers, and investors, AI agents are increasingly being treated not as side experiments, but as high-leverage collaborators that can accelerate thinking, building, and decision-making.
Why this matters
- AI agents are becoming serious infrastructure: Tan’s spending reflects a willingness to pay for tools that may multiply productivity.
- Startups are likely to benefit first: Y Combinator’s ecosystem is built around rapid iteration, making it a natural proving ground for agentic workflows.
- Usage is a strong signal: When experienced builders adopt tools intensely, it often points to practical value beyond hype.
This is not a mass-market breakthrough on its own, but it is a strong sign of where work is heading. As AI agents become more capable and affordable, the kinds of workflows now used by power users like Tan could spread to many more founders, teams, and knowledge workers.