Investors are starting to get jittery about the cost of Microsoft’s push to persuade enterprise customers to adopt expensive generative AI applications and agents.
While the company announced strong quarterly numbers on Wednesday, beating analyst forecasts for both revenue and profit growth, the company’s shares were down close to 6% on Thursday as investors fretted over slightly slower-than-expected growth in its Intelligent Cloud segment, and perhaps more important, sharply rising operating expenses in that segment.
Excitement over Chinese AI startup DeepSeek’s R1 model also rattled investors earlier this week, with some wondering if Microsoft’s decision to base so many of its AI offerings on proprietary models from OpenAI had been shortsighted and whether the tens of billions of dollars Microsoft has been spending to build data centers will be justified if models such as R1 can be run without using as many specialized AI chips.
The irony may be that DeepSeek could help Microsoft CEO Satya Nadella meet high investor expectations around Microsoft’s AI offerings and their ability to drive the growth of its cloud business. While Microsoft’s Azure cloud sales continued to grow at a strong 31% clip year over year, this was less than the 33% growth rate of the previous quarter, and revenues in the overall Intelligent Cloud segment narrowly missed consensus expectations, coming in at $25.54 billion versus analysts’ forecast of $25.76 billion.
This tiny sign of softness may not seem like much, but combined with a 10% increase in operating expenses in the cloud segment and Microsoft also continuing to spend heavily on capital outlays—Nadella reiterated the company’s commitment to spend $80 billion in the next year to build out data centers—Microsoft’s historically robust gross profit margin has begun to shrink. This appears to have spooked some investors.
“Investors and customers are watching closely to see if Microsoft can continue to deliver after the company noted a 70% decline in gross margin for Microsoft Cloud as the result of scaling out AI infrastructure,” Forrester analyst Lee Sustar said.
The cost of deploying AI models within a company, and the uncertainty about finding use cases that will guarantee a return on investment, have held back enterprise adoption of AI across the industry.
Nadella has argued that the innovations DeepSeek pioneered will lower the cost of deploying AI models and that this will result in wider enterprise adoption of AI, not less. That in turn, should benefit Microsoft’s cloud computing business. He also said that OpenAI and Microsoft would continue to find ways to make OpenAI’s models cheaper for customers to run. “If it’s too expensive to serve, it’s no good, right?” Nadella said on a conference call discussing the company’s earnings with financial analysts.
Nadella also announced Wednesday that Microsoft would begin offering DeepSeek’s models through its Azure cloud computing platform. The model has been in high demand from many companies, which see it as a potential way to unlock new use cases where AI can solve tougher business problems or be used as agents—that take actions across other software as opposed to just generating content or making classification decisions.
Because DeepSeek is an open-source model, however, Microsoft’s cloud competitors will also be able to offer it. This may erode Microsoft’s advantage as one of the few options for large business customers to access OpenAI’s models via the cloud. If businesses decide that DeepSeek’s R1 is an equally capable model, Microsoft could find itself competing against other cloud providers in a margin-compressing price war. That prospect may be behind investors’ cooling attitude toward Microsoft’s shares, which had risen 70% since the debut of OpenAI’s ChatGPT in late November 2022.