To protect margins against soaring GPU costs, Microsoft is planning to deploy its custom Maia 300 chip by 2027. It targets a 40% efficiency gain per watt, backed by a multi-gigawatt scaling plan that will require millions of units. This deliberate shift directly threatens Nvidia’s core market among AI hyperscalers.
#japantech — Public Fediverse posts
Live and recent posts from across the Fediverse tagged #japantech, aggregated by home.social.
-
Meta isn't just rivaling OpenAI with its new Muse Wonder AI model—it’s hitting Anthropic where it hurts. By moving internal engineering off Anthropic and onto its own benchmark-topping model, Meta threatens a key enterprise revenue stream right as Anthropic eyes a massive IPO, all while reshaping US tech policy.