View / What goes around, comes around in AI investments

Who captures the value in AI? I increasingly think it’s unlikely to be the creators of the models themselves: Gemini and ChatGPT are currently trading the lead back and forth — Sam Altman, recently lapped, has declared a “code red” — but with enough time, power, and talent, they’ll mostly converge.

It’s more likely to be those upstream (the “picks and shovels” purveyors producing GPUs) and downstream (application makers and corporate customers that can replace workers with AI) of those models. LLMs will be to the AI age what electrons were to the electricity age — critical but mostly fungible, and useless without sockets and things to plug into them. The value lives mostly on the outside of the socket; electric-utility CEOs are not, by and large, billionaires.

In that light, OpenAI’s bizarro-world investment in Thrive — normally venture firms give money to startups, not the other way around — looks like a classic diversification play. OpenAI is worth $500 billion today. But if AI fortunes lie upstream and downstream of it, then it makes sense to put a few chips, so to speak, in those places.

That’s the bearish-on-LLMs case, but the bull case gets you to the same place. If ChatGPT is eventually smarter than human venture capitalists, the VCs will ask ChatGPT what companies to invest in. And now the people who built ChatGPT can own a piece of every company it recommends. It’s a more circuitous monetization route than converting its 800 million weekly users into paying subscribers, but a fun one. Why not tack on an activist hedge fund?