Are you tired? I'm tired. In addition to four Big Tech companies reporting earnings this past week, the calendar included results for second-quarter gross domestic product and the June personal consumption expenditures price index, along with the Federal Reserve's interest-rate decision. Unscheduled, there was a meltdown of the artificial-intelligence trade, as highly leveraged bull positions were liquidated. The iShares Semiconductor exchange-traded fund was down 12% across three days, and one of the most successful AI investors -- hedge fund Situational Awareness -- saw margin calls and forced sales.
But then on Wednesday afternoon, the skies cleared because evidence remains that demand for AI services is growing rapidly and still outstripping supply, even after hundreds of billions of dollars in AI data-center expansion.
Here are my five takeaways from the wild week.
The Capex Is Paying Off
The largest public clouds -- Amazon Web Services and Microsoft Azure -- are on fire. Their results confirmed the success from Google Cloud, reported with Alphabet's results a week earlier.
All three cloud segments handily beat expectations for revenue and, even more crucially, operating margin. As the depreciation expenses from previous capital expenditure pile up, the companies have found operational efficiencies to offset those costs.
The clear message this earnings season is that there is real profit to be had from all of that capex. These companies may be marring their balance sheets and cash flow statements, but their income statements are now seeing a very clear benefit. Despite all of the new data centers already built, cloud capacity remains very tight at all three firms. Amazon.com was even able to raise prices on AI server rentals twice this year, most recently a month ago.
The capex will continue, at least through the end of this year. Google and Amazon raised their 2026 capex guidance, while Microsoft held steady. Together, they will spend about $600 billion this year on AI data centers, and they are hinting at even more next year.
Software Gets a Boost
As nice as Microsoft's cloud quarter was, its best results may have come in business software. Like other enterprise software companies, Microsoft has been under the microscope amid fears that AI will break its sales growth and profitable subscription pricing model. But in the latest quarter, Microsoft's business software segment beat expectations for both sales and operating margin, defying the bear narrative.
Another encouraging sign is that the company's flagship AI software product, Microsoft 365 Copilot, is finally starting to gain traction after a slow start, a common pattern for Microsoft products. The company first revealed paid-user data six months ago -- an unimpressive 15 million out of 450 million Microsoft 365 subscriptions. The latest number is still small, but it has doubled to 30 million.
Salesforce is seeing something similar with its Agentforce AI software, which has seen annual recurring revenue go from $440 million to over a $1 billion in nine months. Software isn't dead yet.
Meta Needs a Cloud
Alone among the big AI spenders, Meta Platforms still doesn't have a cloud unit to directly monetize all of the AI data centers it's building. In its earnings call, the company projected 2026 capex of up to $145 billion -- all for a variety of internal purposes.
Meta sales were up 28% from last year, and some of that performance is due to AI, but mounting depreciation and AI research expenses sent its profit margin plummeting.
The company says that a cloud segment is probably in its future. For now, though, it needs all of the computing capacity internally.
Apple's AI Tradeoff
The supply-chain issues facing Apple suggest that it's no longer the most important buyer of tech components; it has real competition from AI server makers swallowing up output from Taiwan Semiconductor Manufacturing and the memory producers.
Still, for now, the Apple story is largely unchanged by AI. The stock fell 8% on earnings because Apple continues to trade on its iPhone and services segments -- and the latest updates disappointed on both fronts. But the company has plenty of time to get its AI strategy right. The first step will be the release of the new Siri, probably in September.
Meanwhile, Apple hasn't engaged in a mad dash to build AI infrastructure. Though research-and-development expenses are up over 30% this fiscal year, it spent just $2.5 billion on capital expenditure last quarter. Its balance sheet and cash flow statement remain pristine -- an increasing rarity in Big Tech.
Situational Awareness
It's never good news when a tech hedge fund is forced to liquidate its holdings, but the Situational Awareness meltdown holds some helpful lessons, especially for younger investors who haven't lived through a real bear market.
The main lesson is that trading with borrowed money works the same way on the way down as it does on the way up -- you can be right on the fundamentals and still get crushed. Situational Awareness was founded in 2024 by a young former OpenAI engineer with a now-familiar thesis: buy AI hardware stocks and short software. He was absolutely correct, and magnified by margin leverage, the fund's returns were stunning.
But then came July and a rapid selloff for chips. The semiconductor ETF was down 27% from June 30 to July 29, while the iShares Expanded Tech-Software Sector ETF rose by 2%. Margin calls ensued, and Situational Awareness was forced to sell its public-company shares to Ken Griffin's Citadel in a block sale.
There was liquidity for these assets in the end, and that's a good sign. Situational Awareness had just flown too close to the sun. The semiconductor ETF rose 8.5% after the sale on Thursday.
Comments