Goldman Sachs: Microsoft Corp. (MSFT) — Maia 200, Updated Inference Cost Curve, and Internal Silicon Strategy
Summaries are public. PDF access requires an active membership; all members have the same access. Sign in
Goldman Sachs analyzes Microsoft's January 26 introduction of the Maia 200, its updated custom AI inference accelerator. Based on benchmark data and industry feedback, Goldman believes Microsoft has made meaningful progress in its internal silicon strategy. The accelerator's raw computing performance is now broadly comparable to Amazon's Trainium and Google's TPU, representing a positive development for Microsoft's AI compute cost-effectiveness and its long-term goal of aligning Azure AI compute gross margins with CPU workloads.
However, several limitations remain. Maia has not yet been validated in high-volume production environments, and its software ecosystem requires further maturation to support essential developer tools. Additionally, the competitive landscape continues to shift, with Google's TPU v8 expected in 2027.
Goldman emphasizes that diversifying chip infrastructure is crucial for Microsoft to optimize AI returns. This strategy encompasses custom silicon like Maia, increasing adoption of AMD as a viable alternative, and specialized solutions. Goldman expects commercial solutions to maintain the majority market share in the foreseeable future.
Goldman maintains a Buy rating on Microsoft with a 12-month target price of $600. With the stock currently at $397.23, this implies 51.0% upside potential.