No Moat Is Not My Problem

Steve Eisman, the investor from The Big Short, says the AI labs have no moat. He is right. For anyone who builds software with AI, that is good news.

Open-weight models are six to twelve months behind the best models, and the best models from a year ago were already good enough to ship real work with. Most of them do not run on a normal laptop yet, but anyone can host them, so many providers sell them, and they compete on price.

Eisman goes further and thinks the doomsday talk from the labs is a push for regulation that only the biggest players can afford. I cannot see inside their heads. But a company without a moat has every reason to want rules that build one for it.

Most product work is copy changes, forms and normal features, and an older model handles that fine. If token prices double tomorrow, we switch to an open model from a cheaper provider.

So whether AI is a bubble is a question for investors. If it pops, some valuations go to zero, and the open models are still there.