Meta turned down Amodei's personal plea for compute
The chip hunt is back in the headlines — and today's edition runs from a boardroom ask at the very top of the AI industry to the economics of putting robot drivers in truck cabs.
Dario Amodei personally approached Meta earlier this year to source more compute for Anthropic, and Meta declined — according to the Wall Street Journal. The report lands inside the Journal's larger piece on how desperately the industry is hunting for computing power, and it is the kind of detail that only surfaces when rival CEOs are calling each other directly: Anthropic is now among the largest buyers of frontier compute in the world, yet its chief executive still had to go hat in hand to the company he left. The same reporting shows how warped the market has become — Google paid $2.7 billion in 2024 just to bring Noam Shazeer back, after which his team's compute allocation was cut and he departed for OpenAI. Meta and Anthropic are hardly strangers to dealing: earlier this year they discussed a compute lease reported to be worth up to $10 billion over two years, and the New York Times reported in August that Meta internally projected spending up to $10 billion a year on Anthropic's tools. Our take: when a company that might spend $10 billion on your software won't spare you capacity, compute has stopped being a procurement problem and become the industry's actual foreign policy. One caveat — the refusal itself is single-sourced to the Journal so far; it is consistent with everything else reported about the squeeze, but treat it as the Journal's account until a second outlet confirms it.
Is autonomous trucking's revolution here? Phenomenal World's Benjamin Fong says the technology is real but the economics and the regulation are not there yet. Aurora exits 2026 with roughly 200 driverless trucks and an $80 million revenue run rate after logging 500,000 driverless miles, and its pitch to carriers is arithmetic: charge $0.85 per mile for a driver-as-a-service subscription against the $1.03 per mile a human driver costs. But Goldman Sachs only expects around 25,000 autonomous trucks on US highways by 2030, Aurora's stock fell 12.4% after its investor day where it laid out a target of 30,000 trucks by 2030, and Fong's deeper point is that safety is a regulatory question, not a model question — the FMCSA gets under $1 billion and 350 safety investigators a year while the FAA gets $25 billion and over 4,000. With California only opening to AV truck testing in April and no national framework yet, the revolution looks more like a slow freight than a launch.
What to watch: whether a second outlet corroborates the Meta refusal, and whether Aurora's unit economics hold up when its driver-as-a-service model — the higher-margin one — actually scales.
If compute is the real constraint on the AI boom, should labs be able to buy capacity from anyone — even rivals? Tell us in the comments.
