Microsoft's AI balancing act

Welcome to Runtime! Today: Microsoft's second-quarter earnings report pointed to an interesting question about the future of enterprise app development, the fallout from DeepSeek's depth charge continues, and the latest enterprise moves.

(Was this email forwarded to you? Sign up here to get Runtime each week.)

Kind of blue

You know we've come to a weird place in tech history when the price of a company's stock falls more than 6% the day after announcing $69 billion (nice) in quarterly revenue. But such are the expectations that Microsoft set for itself two years ago when it declared that it was the leader of the generative AI revolution, which implied its most important business was about to take off.

Microsoft beat Wall Street's expectations for revenue and profit during its second fiscal quarter, but the company said Wednesday that it expects overall revenue to decline slightly during the third quarter. Compounding the problem, Azure growth appears to have stalled even after Microsoft made changes last year to the products included in the official "Azure and other cloud services" figures that were designed to make Azure growth look better.

• Azure growth came in at 31% for the second fiscal quarter, which is still pretty strong overall but was down from the previous quarter.
• And Microsoft CFO Amy Hood told financial analysts on a conference call to expect flat Azure growth in its third fiscal quarter.
• Microsoft attributed 13 percentage points of Azure's second-quarter growth to AI services, which is up an enormous 157% compared to last year.
• However, if AI's contribution to Azure's growth is skyrocketing but overall growth is down or flat, that means Microsoft is losing ground in non-AI cloud services.

Hood and CEO Satya Nadella acknowledged that problem during the conference call, attributing the slowing growth to continued problems getting new data centers online and "scale motions." That is Microsoftese for sales through partners, a huge part of how enterprise customers buy its software and services.

• It sounds like Microsoft's customers are struggling to balance how they spend their IT budgets on AI workloads versus non-AI workloads, and Microsoft itself is struggling with the best way to incentivize partners after making some changes last summer.
• "As you do that, you learn with your customers and with your partners on getting that balance right, between where to put our investments, where to put the marketing dollars and importantly, where to put people in terms of coverage and being able to help customers make those transitions," Hood said, according to Microsoft's transcript of the call.
• "At a time of platform shifts, you want to make sure you lean into even the new design wins, and you just don’t keep doing the stuff that you did in the previous generation," Nadella said. "You would rather win the new than just protect the past."

As problems go, Microsoft's are the kind you want to have, and Nadella is right that enterprise vendors that focus on the past are on a fast track to irrelevancy. But the discussion points to an interesting question that we won't know the answer to until later this decade: What will be the mix of AI workloads versus non-AI workloads inside your average enterprise?

• Right now, companies are still trying to figure out how to use generative AI in production, and it's hard to imagine that every application under their purview needs an AI touch.
• But Microsoft (and it's hardly alone, to be fair) just spent the last two years telling customers that generative AI was basically the future of everything, and it has incentivized its partners to approach customers that way as well.
• Last month at AWS re:Invent CEO Matt Garman seemed to understand that companies still need to invest in non-AI applications, and tailored his first keynote address in the big chair accordingly.
• As Microsoft heads into the second half of its fiscal year, it sounds like similar discussions are taking place in Redmond.

Rolling in the deep

It's hard to think of anything that has caused more chaos in the AI world of late than DeepSeek. While most would agree the magnitude of Monday's stock-market response was overblown, there's no question DeepSeek's R1 model altered the trajectory of the generative AI narrative.

Cloud companies rushed to provide DeepSeek's model to their customers, including OpenAI's closest partner Microsoft, which The Verge said moved more quickly than usual to add the model to its Azure AI Foundry. Meanwhile, 404 Media won the headline of the week category with its "OpenAI Furious DeepSeek Might Have Stolen All the Data OpenAI Stole From Us" take on OpenAI's lame protests that some of DeepSeek's accomplishments were built on the work of others, which of course is the entire origin story for large-language models in general.

As the week comes to a close it seems pretty clear that DeepSeek spent far more than $6 million to deliver its model, but even the haters had to grudgingly acknowledge that it had made a genuine breakthrough in AI model architecture. Still, a large number of companies blocked their employees from using DeepSeek over concerns about its Chinese origins, according to Bloomberg.

Enterprise moves

Tim Page is the new CEO of Quest Software, joining the IT security and management company from CloudSoda.

Wissam Jabre is the new executive vice president and CFO of NetApp, following several years in a similar role at Western Digital.

Alvina Antar is the new chief digital officer at F5, a newly created position building on her experience as CIO of Okta and Zuora.

Greg Barbaccia is the new CIO of the federal government, joining that train wreck after serving as CISO of Theorem.

Hemant Chaskar is the new CISO of Nile, following several years in related roles at Mojo Networks and Arista Networks.

Daniel Marcu is the new global head of artificial intelligence engineering and science at Goldman Sachs, according to Reuters, joining the company after several years working on Alexa and AI services at Amazon.

The Runtime roundup

The Department of Justice filed a lawsuit attempting to block HPE's acquisition of Juniper Networks, claiming the deal "would eliminate fierce head-to-head competition between the companies, raise prices, reduce innovation, and diminish choice for scores of American businesses and institutions."

ServiceNow missed analyst estimates for revenue and profit and gave a weaker-than-expected outlook for the current quarter, which sent its shares down more than 11% Thursday.

IBM, meanwhile, beat Wall Street's expectations and enjoyed a nearly 13% surge in its stock price Thursday, although it sounds like more cost-cutting is on the way.

Intel, however, continued to struggle even though it beat revenue estimates on a 7% decline in revenue, announcing that it has killed its next-generation AI data-center chip.

Thanks for reading — see you Saturday!