FUNDING

Y Combinator’s Garry Tan Says He Spends Up to $200,000 a Year on AI Tokens

S Sarah Chen Nov 3, 2026 2 min read
Engine Score 7/10 — Important

tier1 source, impact=6, novelty=8, action=5

Editorial illustration for: Y Combinator's Garry Tan Says He Spends Up to $200,000 a Year on AI Tokens
  • Y Combinator CEO Garry Tan told podcaster Walter Isaacson he spends “$100,000 to $200,000 a year on tokens” running AI agents.
  • Tan has publicly shared a set of 23 Claude Code skills on GitHub, called gstack, to show others how he uses agents.
  • Y Combinator built an internal agent named Stray Light that holds the full context of every founder the accelerator has backed.
  • Tan says personal AI agents won’t go mainstream until token costs come down, calling his own spending “a crazy privilege.”

What Happened

Y Combinator CEO Garry Tan told Walter Isaacson, in an episode of Isaacson’s podcast The Builders set to air Tuesday, that he spends “$100,000 to $200,000 a year on tokens” running AI agents (TechCrunch, Nov. 2, 2026). Tan said he uses agents for work projects, coding tasks, and managing his personal life.

Why It Matters

Tan’s spending illustrates the real cost of heavy agentic AI use, a topic MegaOne AI has tracked through token-budget breakdowns for Claude Code users and the rise of AI-written code on GitHub. Tan is an early OpenClaw adopter and has shared 23 Claude Code skills, called gstack, on GitHub to show other users what agents can do.

Technical Details

Tan said his agent has read “all of my email and all of my Dropbox,” along with his Apple Notes, including therapy notes going back 20 years, to act as a personal assistant. Y Combinator separately built an internal agent named Stray Light that Tan says holds “the total context of every founder we’ve ever backed” and everyone who has applied to or emailed the accelerator.

Who’s Affected

Startup founders and YC applicants are directly affected, since Stray Light processes their application and communication history. Heavy individual users of coding agents like Claude Code face similar token-cost tradeoffs to the one Tan describes, even at far smaller scale.

What’s Next

Tan said mainstream adoption of personal AI agents depends on token costs falling, calling his own spending “a crazy privilege that not everyone can do.” He did not give a timeline for when he expects agent costs to decline.

Share

Enjoyed this story?

Get articles like this delivered daily. The Engine Room — free AI intelligence newsletter.

Free · No spam · Unsubscribe anytime