Precarious Times

This has gotten lots of attention in tech circles, but not that many people have been connecting the dots. 

From the Information: 

Coinbase has cut its AI spending “nearly in half” even as it increases the number of tokens it uses, by using various measures to control costs. These measures include defaulting to open-weight models from Chinese firms, Coinbase CEO Brian Armstrong said in an X post on Friday night.

Armstrong’s post highlights both the growing usage among U.S. firms of open-weight models coming out of China to control AI spending even as the cost of cutting edge U.S. services rises. Armstrong said Coinbase had experimented with GLM 5.2, a just-released model from Z.ai, and Kimi 2.7, from Beijing-based Moonshot.

Pretty much everybody reading this has a good idea how massive the AI boom is, with current and proposed capital expenditures in real terms exceeding those of the railroad bubble, the dot-com bubble, and the early-2000s real estate bubble combined. And anyone who seriously follows the markets knows the narrative that drives the bubble, one of fantastic profits and unprecedented gains in productivity. But there's one aspect of that narrative that doesn't get nearly the attention it deserves: in order to justify the valuations of these companies and the capex being spent, you also have to assume that this will be a high-margin industry primarily based on one or more of the big, proprietary frontier models running on enormous centralized data centers and owned by either OpenAI, Anthropic, Meta, Google, or xAI.

If a substantial part of the industry follows the lead of Coinbase, the narrative is suddenly very different, with serious, even catastrophic, ramifications for some of the biggest, most powerful companies in the world.

Anthropic, to a degree, is protected by a moat here. Its standing as the leading model for coders protects it somewhat, but for OpenAI, which is already struggling with multibillion-dollar losses, a pushed-back IPO, and hundreds of billions of dollars of commitments coming up in the next three or four years, increased competition and substantially lower profit margins could present an existential threat.

Some of the biggest companies in the world have a high degree of exposure to OpenAI. The debt-ridden Oracle would be in serious trouble if anything happened to the company, as would SoftBank. Chip makers, including the most highly valued corporation in the world, Nvidia, would easily weather the collapse of OpenAI, but they would take a substantial hit in terms of revenue and almost inevitably stock price.

We won't even get into the possible ripple effect that might hit the private credit market.

I'm not saying any of this is likely to happen, but it seems worth thinking about, particularly if you've got money in the tech sector.