Every major tech CEO in America is committed to the AI capex arms race. Meta has raised its 2026 capital budget to between $125 and $145 billion. Alphabet has guided to $180 to $190 billion and says 2027 will go higher. Microsoft is on track for roughly $190 billion this year.
Brian Chesky has told Wall Street, repeatedly and proudly, that Airbnb will do none of it: no model training, no GPU buildout, no AI investment the market will ever see in the P&L. Airbnb will stay as the application layer, the disciplined opposite of the hyperscalers, winning on what it does with commodity models rather than on building them.
Three weeks after his most recent earnings call, it came out that Chesky was starting a new AI frontier lab — with a mix of personal funds and outside investors — to build models, separate from Airbnb, run by someone other than him. Which of course is a contradiction because Chesky built his entire investor story on not building models, and i
source
