Broadcom's Hock Tan, the Shovel Seller, Calls AI Investors Gamblers

The company that sells the shovels stood in front of a room of gold miners and told them they were gambling. Broadcom's Hock Tan said at a Singapore conference that many AI investors "have no clue what they are investing in" and set "very extreme expectations only to find out that it's almost like they are in a casino." Then he kept his own AI revenue guidance: about 58 billion dollars this year, 230 billion by fiscal 2028.

Tan's warning and Tan's guidance can both be true. Tan sells custom chips to a few hyperscalers on multi-year contracts. Those contracts fill Tan's order book, and they are the last spending anyone cuts. He aimed his warning at everyone downstream of that order book.

Tan's gamblers match the gambler I described in AI is Magic, who sits below the wizard. The gambler in that post has no ritual and no method, and stakes the outcome on pure chance. An investor with "no clue" what the money buys makes the same bet on a mechanism that the investor cannot audit. Some of those bets end badly, and a gambler has no method to repeat.

Why would capital become scarce before chips?

Tan's third warning on the same panel matters more than the casino line. He said that capital may become scarce sooner than chips. Outside money pays for the buildout. Companies pay for data centers with debt raised against revenue that often does not exist yet. Tan pointed to one firm, Firmus, which pulled its IPO the week before. If capital becomes scarce, orders stop before chips become scarce.

Funding stress appears first in deals that quietly do not happen, before any headline. The Firmus IPO is one such deal. Tan, who sees his order book from the inside, did not cut his guidance and told you to watch the deals that do not happen.

Is the risk that the models disappoint, or that the money stops before the models deliver?