Microsoft FY 3Q'26: Multi-Model Mirage, Copilot Momentum

Key Takeaways

• 10,000 Foundry customers used multiple AI models, 5,000 used open source
• Multi-model adoption still low; ~90% of customers use a single model
• Copilot usage up 6x YTD; weekly engagement matches Outlook
• Seat‑based pricing bundles consumption, with overage discounts for committed usage
• Microsoft relies on royalty‑free OpenAI model until 2032, lacks own frontier model

Pulse Analysis

Microsoft’s latest earnings reveal a nuanced picture of enterprise AI adoption. While the company touts a multi‑model strategy, the numbers suggest the majority of its 80,000 Azure AI Foundry customers remain anchored to a single provider. This mirrors broader enterprise tendencies to standardize on one cloud or AI stack to avoid the operational overhead of managing multiple vendors. The modest growth in customers using both OpenAI and Anthropic models—doubling to roughly 3,000—highlights that multi‑model adoption is still in its infancy, leaving Microsoft with a long runway to convince CIOs of the strategic benefits.

The real headline from the call is Copilot’s rapid uptake. Monthly active users have surged six‑fold year‑to‑date, and weekly engagement now rivals that of Outlook, Microsoft’s flagship productivity app. This momentum underscores the growing reliance on AI‑augmented workflows across the enterprise and positions Copilot as a potential revenue catalyst for Azure. The company’s seat‑based pricing model, which bundles a baseline of consumption with overage discounts for committed usage, aims to provide cost predictability while still capturing high‑volume token usage—a balance that could appeal to budget‑conscious IT departments.

Looking ahead, Microsoft’s strategic reliance on a royalty‑free OpenAI model through 2032 raises questions about its ability to develop a proprietary frontier model. Without a differentiated offering, the firm may struggle to maintain a competitive edge as rivals like Amazon and Google accelerate their own model development. Additionally, as the current "token‑maxxing" phase matures, enterprises will likely shift focus from experimentation to cost optimization, potentially compressing AI‑related margins. How Microsoft navigates these dynamics—balancing multi‑model aspirations, Copilot growth, and pricing flexibility—will be pivotal for its AI revenue trajectory.