AI Capex Pressures Microsoft and Oracle Margins

Microsoft’s own filings show how quickly the spending is accelerating. In the nine months ended March 31, it said cash used in investing jumped $42.6 billion to $84.7 billion, driven largely by additions to property and equipment, as well as components tied to the buildout. It also acknowledged that gross margin percentage slipped because of continued investment in AI infrastructure and rising AI product usage. In other words, the scale of demand is real, but so is the cost of serving it.

Oracle is running the same playbook, only with more obvious balance-sheet risk. The company has leaned on borrowings to fund its expansion, and that kind of capex can be a powerful growth engine when utilization stays high. But when debt rises faster than operating efficiency, margin compression tends to show up before the payoff does. Oracle’s stock, which has been wildly volatile, reflects that tension: investors like the AI and cloud opportunity, but they are also recognizing that a leverage-heavy buildout can leave less room for error.

The price action tells the story. Microsoft has fallen to about $381.70 from a recent high near $460.52, while Oracle has dropped to $114.99 from above $235 in June. That kind of decline is not just about technical weakness — both names are now trading well below their recent peaks, with Microsoft below its 200-day moving average and Oracle far beneath it. Conventional indicators such as RSI also point to heavy selling pressure in both stocks, which usually happens when sentiment shifts from “how big is the AI opportunity?” to “how expensive is it to capture?”

That shift matters for investors because capex intensity changes the whole investment case. A company can absolutely create enormous value by building data centers, chips, and cloud capacity ahead of demand. But if it finances that expansion with debt and the return on that capital takes too long to arrive, margins get squeezed and the market can stop paying up for growth. This is especially true when broader risk appetite cools, as shown by the fearful tone across the S&P 500 in Adalytica’s trade-signal snapshot.

The long-term thesis for Microsoft and Oracle is not broken. Both companies have real moats, deep enterprise relationships, and exposure to one of the most durable growth trends in technology. But the next few quarters may look less like a straight-line AI payoff and more like a necessary digestion phase, where spending outruns earnings and investors reassess how much they want to pay for future optionality.

For patient investors, that may create opportunity rather than danger. The AI infrastructure buildout is likely to last for years, not months, and the strongest platforms should eventually compound through it. But the winners will be the companies that turn capex into operating leverage, not just into bigger balance sheets. Microsoft and Oracle remain worth watching, but this is a moment to focus on cash generation, not just AI enthusiasm.