Stream of glowing tokens flowing into an AI agent core on a dark navy background

YC Boss Spends Up to $200K a Year on AI Agent Tokens

In short

  • Y Combinator CEO Garry Tan says he spends about 100,000 to 200,000 US dollars a year on tokens for his AI agents.
  • He called it a privilege most people cannot afford, and said personal agents go mainstream once costs fall.
  • That bill equals the salary of two to four developers. Your own agent budget should be a fraction of it.

One of the most powerful people in startups just told you how much a heavy AI agent user pays. It is a lot. The useful question is what you get for that money, and what you should spend instead.

What happened

On Walter Isaacson’s podcast The Builders, Garry Tan said he spends about 100,000 to 200,000 US dollars a year on tokens, according to TechCrunch (published November 2, 2026). He runs agents on many work projects and on coding. He published a set of 23 Claude Code skills called gstack on GitHub so others can copy his setup.

He also said Y Combinator runs an internal agent called Stray Light, with context on every founder the company has backed and every application it has received.

His view: agents should not be treated like traditional software, and too much focus on repeatability and safety throws away what makes them useful. He expects personal agents to go mainstream when costs come down, and he prefers a personal AI you control and program over free agents run by Big Tech.

How to use it

Do the math: 100,000 to 200,000 dollars a year is about 8,300 to 16,700 dollars a month. At roughly €0.90 per dollar (my approximation), that is about €90,000 to €180,000 a year. A developer at €45,000 a year costs a half to a quarter of that bill. So Tan pays the price of a small team, for one person with agents.

You do not need that. A solo founder with a €200 monthly agent budget spends €2,400 a year, which is 1/40 of the low end. The lesson is not to copy his bill. Copy the method: give the agent real context about your business, and put it on work that repeats.

Start with one workflow that costs you hours each week: replying to customers, preparing orders, checking listings. Track the token cost against the hours saved. If one hour of your time is worth €30 and the agent saves ten hours a month for €40 of tokens, you are €260 ahead.

What it means

Token spend is becoming a real line in the budget, like payroll. The people who learn to measure it early will use agents where they pay back and cut them where they do not.

Keep one thing in mind. This is one person’s statement in an interview, not an audited number, and Tan is an enthusiast with a lot of money and a large platform. His use is extreme by design.

Ivan’s Take

I do not care about the headline number. I care that someone at that level sees agents as a payroll-size cost and still pays it. I would do the same at a smaller scale: pick one workflow, measure tokens against hours, and scale only what pays back.

Source: TechCrunch

II
About the author
Ivan Ivanov
Founder ofAI Rush XGiantPartsSupreme CommerceHawkValue

If it doesn’t save you time or money, you won’t see it here

  • Facts, not hypeEvery price and number is checked against the official source
  • Always the real priceYou see what it costs before you click
  • I say when it’s weakEven when it’s OpenAI or Google
All articles by Ivan →

Similar Posts