The Arms Race Is Not a Metaphor

China is reportedly committing $150 billion in state-backed capital to AI. I haven't confirmed every number — this is moving fast enough that sourcing has a half-life of about six hours — but the broad shape of it is almost certainly true, because it's the obvious move.

Five days ago DeepSeek dropped R1, which was notable partly because it matched frontier performance at a fraction of the compute cost. The story everyone ran was: China can do more with less. The thing nobody said clearly enough is that "doing more with less" and "also spending more" are not mutually exclusive. They're actually the combination that should scare you. Efficiency plus scale is how you win wars, literal or otherwise.

$150 billion directed at a specific technology by a single coordinating entity is not a market signal. Markets don't do this. This is industrial policy — the same playbook that built TSMC in Taiwan and the Korean chaebols — applied to a sector that everyone has now decided is the terrain of the next forty years of geopolitical competition.

The arms race framing is a cliché at this point, which usually means it's accurate. We spent decades arguing about whether AI was really transformative or just a fancy autocomplete. That debate is over, not because anyone won the argument, but because governments started treating it like a strategic asset — and governments have a way of making their assumptions self-fulfilling.

Nobody is asking whether this is healthy. That ship sailed.