Remarkably little economic value sits beneath the massive AI capital expenditure backlog. The net present value of future rental payments—after deducting the capex required to provide the compute—equals
Four things worth thinking about this evening: 1. South Korea leverage unwind. The following Scooby Doo statistic made the rounds today, and, while an exaggeration, it is directionally useful in
Something I wrote recently in another context and shared with some friends. Its inherent homage to clergyman Jonathan Swift will be obvious. Have a good weekend. Paul A Modest Proposal
Chinese large language model usage among U.S. companies (on OpenRouter) has soared in the first half of 2026. After being in the low single digits through early 2025, token
Falling birth rates shrink the workforce and slow economic growth. This has been economic orthodoxy since Adam Smith. New research finds the (almost) opposite, however. Countries with larger baby busts
The U.S. labor force participation rate has quietly returned to levels last seen in the 1970s (outside the recent pandemic). It increasingly seems to signal a structural shift in
A quick roundup of five things I'm thinking about and have been discussing with clients and others lately. 1/ Taiwan is now the world's fifth-largest
Chinese patents are mostly built on Chinese science rather than American science, according to a new working paper. Across every "critical technology area" shown, the U.S. share
I did a lengthy interview a little while back, and it is now out. Folks may find it a helpful restatement of some of my views on the current situation,
I have a short essay in the NY Times today on why Americans are such global outliers in their AI antipathy. Hating AI is, as I say in the piece,
China did to solar panels what everyone said couldn't happen to a complex manufactured good: it turned them into a commodity, drove prices down 90%, and eliminated the
Models are getting (slowly) better, which will make things (quickly) worse
