Original source：Morgan Stanley & Co. LLCFirst published on Unhidea：Original source date：
Key takeaway
Morgan Stanley is optimistic about Microsoft’s advantages in infrastructure, model management and enterprise distribution in a multi-model era. The key is whether cheaper inference can generate enough additional usage to offset pressure on revenue per inference and gross margins.
Why might the rise of open models benefit Microsoft?
Among the top 50 models by usage on OpenRouter, open-weight models’ share of tokens rose from about 22% a year ago to about 72%; since the start of 2026, the number of Microsoft customers developing with multiple model providers has increased 5-fold.
Morgan Stanley believes enterprises will continue to choose models based on cost, performance, latency and data sovereignty requirements. Foundry supports more than 11 thousand models and provides a unified layer for access, management, evaluation and deployment; Azure provides the underlying infrastructure for workloads.
Microsoft’s portfolio also includes Fabric, which provides enterprise data context, and M365 and Copilot, which bring AI capabilities into workflows. More than 17 thousand customers currently use both Fabric and Foundry, up 60% year on year.
How can existing workplace access points drive AI usage?
M365 and Copilot are already embedded in enterprise workflows such as productivity, collaboration and communications, providing ready-made distribution channels for AI agents. M365 Copilot currently has more than 30 million paid seats.
Morgan Stanley’s CIO survey shows that 47% of surveyed CIOs prefer Microsoft for building custom AI applications, versus 10% each for AWS and Google. The bank believes that lower inference costs will allow more high-frequency, low-value tasks to be automated, giving Microsoft an opportunity to extend these capabilities across its existing user base and drive application and Azure consumption.
How do falling inference prices affect revenue and costs?
Lower inference costs could reduce revenue per inference, but could also bring AI to more workflows. Morgan Stanley’s view depends on whether increased usage can offset the decline in revenue per inference.
Microsoft is expanding usage-based billing: for example, Copilot Studio uses Copilot Credits to charge based on agents’ actual consumption. A multi-model architecture also allows Microsoft to assign tasks to suitable models based on cost and performance, thereby optimizing the cost of providing services.
How are growth and margin expectations reflected in the valuation?
Morgan Stanley expects revenue growth for Azure and other cloud services to rise from 40.7% in fiscal 2026 to 44.9% in fiscal 2027, before easing to 43.2% in fiscal 2028; over the same period, the company’s overall gross margin is expected to fall from 67.9% to 64.5%.
The bank expects operating leverage to partly offset gross margin pressure, with an operating margin of 46.7% and earnings per share of 24.08 dollars in fiscal 2028. Increased investment, limited AI adoption or a macroeconomic drag on IT spending are downside factors for its rating and price target.
Morgan Stanley maintains its Overweight rating and 600-dollar price target, based on a price-to-earnings multiple of 25 times its fiscal 2028 earnings-per-share forecast; the closing price listed in the report for 23 October 2026 was 524.51 dollars.
