Big Tech Is Spending Gazillions on AI. Earnings Offer a Peek.

This past week’s batch of earnings reports from companies such as Meta Platforms, Microsoft, Intel, Alphabet, IBM, and ServiceNow provided a slew of insights—some more obvious than others—about what just happened, and where we go from here. Here are the key takeaways.

• The Clouds Are Thickening. Almost all of what I’m about to tell you ties back to this single development: The cloud titans are spending gargantuan sums to build out AI data centers, and it seems to be paying off. Microsoft reported 31% growth in its Azure Cloud business in the March quarter, about three percentage points higher than the Street had expected, with seven points of growth coming from AI-related workloads. The company expects June-quarter growth to be in the 30% to 31% range, ahead of previous estimates. Meanwhile, Google Cloud grew 28% in the quarter, two points ahead of Street estimates.

When you have that kind of growth, you feed the beast. Microsoft’s capital spending in the March quarter was $14 billion, and the company expects the numbers to rise from there. The implication is that Microsoft’s capex in the June 2025 fiscal year will easily blow past $50 billion. According to FactSet, the only other U.S. public company on track for 2024 capital spending at that scale is Amazon.com, Microsoft’s primary cloud rival, which reports results next week. Google is no shirker either, spending $12 billion in the March quarter. The company told investors that the figure will be at least that level or higher for every quarter this year, which suggests another $50 billion-plus.

Meta told investors that its 2024 capital spending could be as high as $40 billion—and warned that it could take several years before those outlays result in significant related revenue. That spurred a roughly 10% selloff in Meta shares. Unlike Microsoft and Google, Meta has no public cloud business, which seems like a strategic error. It certainly would have been a better bet than the Metaverse. But the point is clear: AI computing is a game only for the biggest of the big.

To put this in a little perspective, compare what’s going on in AI to what had been the world’s most capital intensive industry—the oil business. Saudi Armaco, by far the industry leader, is expected to have capital spending this year of $50 billion, about equal to Microsoft. Meta’s capex budget will be about equal to Exxon Mobil and Chevron combined. Yep, data is the new oil, in more ways than one.

• Hardware Is Eating the World. Remember Marc Andreessen’s declaration that “software is eating the world?” Not anymore it isn’t. Hardware is eating software for lunch. All that capital spending is going to buy AI servers powered by AI chips, associated storage and networking infrastructure, and the required cooling and power. The $200 billion or so that the four largest cloud spenders will shell out in the next year is going to pour into the pockets of chip companies like Nvidia, AMD, Micron and Broadcom; networkers like Arista Networks, Ciena, and Cisco Systems; and server companies like Dell Technologies, Hewlett Packard Enterprises, Super Micro Computer, and Lenovo Group. It’s also good for cloud software vendors like Snowflake, MongoDB, Datadog, and newly public Rubrik —anyone that manages, massages, and protects data. There’s a sellers market for AI hardware; even Microsoft said it has more AI demand than its infrastructure can support. Which is why Microsoft and its peers are out shopping for more, more, more.

• Software Awaits the Payoff. The wild card—the issue to watch in the quarters ahead—is when enterprise software companies start to show measurable progress in driving revenue with AI. Oh, everyone talks about it. ServiceNow CEO Bill McDermott told Barron’s last week that “every business workflow will be re-engineered with Gen AI.” IBM CEO Arvind Krishna told me his firm now has a book of business for AI software that tops $1 billion. But even Microsoft so far is mostly demonstrating a big boost from AI for cloud demand—just how and when the application side of the business will benefit is less clear.

• Everything Else Is Meh. IBM shares tumbled on weakness in its consulting business. Krishna told me companies are hesitating to launch discretionary projects in a higher-for-longer interest-rate environment. Intel CEO Pat Gelsinger told me that the near-term chip outlook is soft, with customers moving cautiously on everything that doesn’t involve AI. But Gelsinger is optimistic about the second half, when the company starts shipping not only new processors for AI PCs, but also AI server chips to take on Nvidia. Like everyone else right now, Gelsinger’s vision for Intel is simply this: “AI everywhere.”

Write to Eric J. Savitz at eric.savitz@barrons.com