Sources: Microsoft is helping finance AMD's AI chip expansion and working with the chipmaker on Microsoft's own processor for AI workloads, codenamed Athena
and its own slate of newly introduced AI services — are requiring computing power at a level beyond what the company expected when it ordered chips and set up data centers.” https://www.bloomberg.com/... Ben Bajarin / @benbajarin : So it's semi-custom. Semi-custom is the new fully custom of the data center. https://twitter.com/... Dylan Patel / @dylan522p : Microsoft Athena is allegedly AMD, according to @business. Surprisingly given there was some chatter, it was GUC backend. It's H1 2024, so between MI300 and MI400. With that said, it's possible. $MSFT $AMD $NVDA @cyw60 https://twitter.com/... Dina Bass / @dinabass : Microsoft is working with AMD both on Microsoft's own AI GPU, code-named Athena, and on bolstering AMD's own efforts to become a second supplier of GPUs apart from Nvidia, sources tell @ianmking and me: https://www.bloomberg.com/...
Context & Ripple Effects
This report builds on the earlier disclosure that Microsoft had been developing Athena and testing it with some Microsoft and OpenAI staff, turning an internal chip effort into a reported co-development and supply-financing relationship with AMD. Microsoft's prior semi-custom processor work with AMD for Surface provides a narrower precedent for the partnership model.
The later plan to make AMD's MI300X available through Azure shows why a Microsoft-AMD alignment mattered beyond one internal chip: it could support both Microsoft's own infrastructure and its cloud hardware menu.
First-order effects
- Microsoft gains a reported route to expand AI compute supply through AMD while shaping Athena around its own workloads, rather than relying solely on an external GPU roadmap.
- AMD receives reported financial support and a major cloud customer relationship as it tries to establish itself as an alternative AI GPU supplier to Nvidia.
Second-order effects
- Microsoft can use a closer AMD relationship to diversify AI-chip procurement and potentially align Azure capacity planning with AMD's product development.
- Nvidia faces a better-resourced rival for large cloud deployments; other chip suppliers must compete not only on hardware performance but on willingness to support customer-specific designs.
Third-order effects
- If cloud operators continue to co-finance and co-design accelerators, AI compute could shift toward a more heterogeneous supply base in which hyperscalers influence chip roadmaps directly.
- That model can deepen vertical integration between cloud platforms and silicon partners, while leaving the durability of any Nvidia alternative dependent on AMD's execution and software adoption.
The trend: Hyperscalers are moving from buying standard AI accelerators to shaping multi-supplier, semi-custom compute stacks around their own infrastructure needs.