Microsoft plans to use Intel's 18A tech to make a forthcoming chip that Microsoft designed in-house; Microsoft recently announced two computer and AI chips
Ian King / Bloomberg :
Context & Ripple Effects
Microsoft’s use of outside silicon had already been diversifying: it was reported to shift much of Azure’s coprocessor demand toward Xilinx-based coprocessors rather than rely exclusively on Intel. By 2020, the company was reported to be developing its own cloud-server and Surface chip designs.
The announced foundry plan gives that design effort a prospective manufacturing route. It follows Microsoft’s November disclosure of Maia for Azure AI workloads and Cobalt for servers, bringing its first named AI and server chips into view.
First-order effects
- Microsoft gains Intel 18A as a planned fabrication option for a forthcoming internally designed chip, extending its control over chip design without requiring it to manufacture the chip itself.
- Intel gains a prospective high-profile foundry customer for its 18A technology, while Microsoft’s recently announced AI and server-chip program gains an additional supply-chain path.
Second-order effects
- Microsoft can more directly align future silicon choices with Azure and AI workloads, potentially reducing the degree to which its infrastructure roadmap is dictated by off-the-shelf processor availability.
- The arrangement raises the competitive importance of foundry access for cloud companies pursuing custom chips: chip designers will weigh manufacturing partners alongside architecture and software fit.
Third-order effects
- If similar arrangements persist, cloud infrastructure will increasingly be built around heterogeneous, purpose-designed silicon rather than a single general-purpose processor supplier.
- The durable strategic divide will be between companies that can pair proprietary chip designs with reliable external manufacturing and those that remain dependent on merchant silicon; the outcome will depend on execution and production availability.
The trend: This is one data point in the shift toward cloud providers treating custom silicon and foundry relationships as strategic AI-infrastructure leverage.