Two of the biggest names in artificial intelligence are bumping up against the same hard limit: there isn't enough computing power to go around.
According to the Financial Times, citing sources, Google told Meta around March that it could not offer all of the Gemini capacity that Meta wanted to buy. Gemini is Google's family of advanced AI models, and Meta — the parent of Facebook and Instagram — had been looking to purchase more of that capacity than Google was able to provide.
The Financial Times reports that the shortfall disrupted and delayed some of Meta's internal AI projects. In other words, even a company with Meta's resources found itself waiting because a supplier could not meet its order.
What's driving the squeeze is demand. The Financial Times frames the episode as a sign that surging appetite for advanced AI models is turning computing power into the tech industry's scarcest commodity. As more companies race to build and run sophisticated models, the underlying hardware and capacity needed to do so are not keeping pace.
The detail that Google declined to sell a rival all the capacity it sought is also notable in itself. It shows how the firms competing fiercely in AI are at the same time entangled as customers and suppliers to one another.
Why it matters: when computing power becomes the bottleneck, even the wealthiest tech giants can have their AI ambitions slowed — a constraint that could shape which products reach users and when.