Microsoft shares jump after report says it is routing Copilot queries to in-house AI instead of OpenAI and Anthropic

Microsoft’s stock rose on Tuesday after a market report said the company has begun routing some Copilot queries inside major Office applications away from outside model providers such as OpenAI and Anthropic and toward Microsoft’s own proprietary AI models.

The report, carried by Yahoo Finance and syndicated through Stocktwits, said the change is part of an effort to reduce the cost of running AI features at scale. It framed the development as an operational adjustment, suggesting Microsoft can improve margins by relying less on paid third-party model capacity for certain classes of requests.

Copilot, Microsoft’s umbrella name for AI-assisted features embedded across products like Word, Excel, PowerPoint, Outlook, and Teams, generates responses by interpreting user prompts and using machine-learning models to produce text, summaries, or draft content. When those models are hosted by external vendors, usage-based costs can rise quickly as engagement grows.

The report did not provide details on which Copilot prompts would be affected, how the routing decisions are made, or whether the change is limited to specific tenants, languages, or customer tiers. It also did not explain whether Microsoft’s own model options are being selected automatically, whether users see any difference, or how performance is being evaluated.

In the absence of a Microsoft confirmation in the published post, the claim should be treated as unverified. Microsoft has not, in the information provided here, described any formal change to its external-provider strategy, pricing, or model mix for Copilot in Office.

Still, the reported direction fits the broader pattern in enterprise AI, where large platform providers seek to manage compute expenses by balancing multiple model sources. For Microsoft, Copilot is delivered through its cloud and productivity ecosystem, meaning small changes in model economics can matter because usage is potentially high and latency-sensitive.

The market reaction also underscores how investors are watching the unit economics of AI features, not just revenue headlines. A shift toward internal models, if it is occurring at scale, could reduce marginal costs per query, potentially helping Microsoft’s profitability as Copilot demand grows.

What remains unclear is whether Microsoft’s in-house routing fully replaces OpenAI and Anthropic models or simply reallocates some portion of traffic. The next datapoints to watch are any official disclosures from Microsoft about model partners, cost structure, or changes to Copilot capabilities and billing, including whether there are measurable impacts on margins or guidance.