The reported refusal makes those prospective capacity discussions look selective rather than purely transactional. Meta's earlier internal consideration of competitor models also shows how quickly nominal model rivals can become infrastructure counterparties when compute is scarce.
First-order effects
Anthropic is denied the additional Meta capacity Amodei sought, leaving Meta in control of whether—and on what terms—it supplies a leading model rival.
Meta and Wang become direct gatekeepers in a negotiation over a resource Anthropic needs to train and serve AI models.
Second-order effects
The reported Meta-Anthropic rental talks gain added strategic weight: capacity access is a bilateral bargaining issue, not simply a commodity purchase.
Other model developers seeking external compute face a market in which infrastructure owners can distinguish between customers and competitors, increasing the value of direct executive relationships.
Third-order effects
If selective capacity sharing persists, AI competition will be shaped not only by model performance but by which developers can secure infrastructure agreements with rival platform owners.
The pattern points toward compute as strategic leverage: companies that own large data-center capacity can influence competitors' operating options without changing their own model strategy.
The trend:AI infrastructure is becoming a strategic bargaining asset, pushing model rivals into selective supplier-customer relationships.
AI startups can raise billions and still struggle to get enough computing power. Even rivals are now making deals to get the chips they need. Money alone is no longer enough. For AI startups, access to computing power could become as important as funding.
https://www.wsj.com/... Compute is the ultimate moat. No one has more than $META. Dario called @alexandr_wang earlier this year trying to source chips. Meta debated sharing and decided not to “for now.” When rivals are begging at your door, YOU are the gatekeeper. via @wsj
Demand for data centers is pushing up hourly cost of renting Nvidia's H100 chips. A one-year contract has risen 60% in the past year. Falling AI and data center revenues but shortages of chips, and rumors of chips in warehouses? Hoarding? https://www.wsj.com/...?
WSJ is reporting that Anthropic CEO Dario Amodei personally called Meta's Alexandr Wang seeking more compute. Meta considered providing chips, then declined. Inside GOOGLE, Demis Hassabis has expressed frustration about Google not allocating enough compute to long-term research. …
“If you overbuy, you could bankrupt the company by having too many lease commitments, and if you underbuy, you could knock down your growth trajectory and have a bunch of unhappy users,” said @thezbg, a general partner at @GradientVC https://www.wsj.com/...
Compute is the lifeblood of the AI boom, and it's in short supply. Now fights over how to get more of it—and what projects to give priority to—are sending a powerful industry scrambling. https://www.wsj.com/... via @WSJ