Microsoft is pivoting its infrastructure strategy, revealing margin pressures behind its AI-driven revenue growth.
An uncomfortable question is forming at the heart of Microsoft’s AI strategy, one unrelated to demand. The company is winning the race to supply computing power to virtually everyone who needs it. Reports indicate Microsoft is courting Chinese rival Moonshot and its Kimi K3 model with a revenue-sharing arrangement to host on Azure. Meanwhile, Meta is rumored to rank among Azure’s largest AI customers, and a partnership dubbed “HUMAIN” is taking shape in Saudi Arabia. The message to investors is unmistakable: Microsoft is transitioning from a software vendor to the landlord of the AI era, renting infrastructure to competitors and partners alike.
The market rewarded this vision on Friday with a 2.2 percent gain to EUR 443.25, extending a seven-day advance to 7.1 percent and a 30-day rally to a striking 30 percent. Notably, this run has been driven largely by quarterly results released roughly a month ago rather than fresh catalysts. On July 29, Microsoft reported fourth-quarter revenue of $90.01 billion, up 17.75 percent year over year, with adjusted earnings per share of $4.74—nearly 12 percent above the consensus estimate of $4.2397. The stock jumped 15.51 percent that day, marking its strongest one-day reaction to earnings in six quarters.
Cloud performance dominated the headlines. Azure grew 43 percent, while commercial remaining performance obligations climbed to $678 billion, an 84 percent increase. Management guided to 45 percent Azure growth in constant currency for the current quarter, signaling acceleration rather than deceleration. This is bolstered by an extended OpenAI partnership through 2032 featuring a $250 billion Azure commitment, alongside more than 30 million paying Microsoft 365 Copilot users. Fundamentally, the story appears formidable.
Yet the pivot to infrastructure carries a structural price that analysts are beginning to flag. Seeking Alpha has raised valuation concerns, highlighting a decline in Microsoft’s cloud gross margin from 72 percent in 2024 to 65 percent this year. Renting out compute capacity simply does not generate the same margin per dollar as selling high-margin software licenses. Investors must now weigh whether the sheer volume of AI-driven demand can offset this erosion long-term. The pressing question remains: should investors sell immediately, or is buying Microsoft still warranted?
Cost pressures are emerging in unexpected areas. An internally leaked table reportedly revealed that a single employee generated $28,000 in AI usage costs within 28 days, sparking internal debate over “tokenmaxxing”—the artificial inflation of usage metrics. While seemingly minor, this detail underscores a critical reality: the AI boom generates both revenue and cost risks that can directly bleed into margins.
Despite these headwinds, investment activity remains aggressive. Microsoft recently approved the first Horizon data center for IREN under a multi-year infrastructure agreement valued at $9.7 billion, expanding Azure’s AI training and inference capacity. The company is clearly betting that compute demand will persist for years, preferring to hold excess capacity rather than risk supply shortfalls. However, the chart presents a more cautious picture. The relative strength index sits at 72.3, indicating overbought conditions, with the stock trading roughly 18 percent above its 50-day moving average and about 20 percent above its 200-day average. Such deviations typically resolve through a pause or correction rather than smooth consolidation. At EUR 443.25, shares remain 7.3 percent below their 52-week high of EUR 478.10 set on October 28, 2025, leaving room for upside without requiring new records.
The capital burden is substantial. Microsoft plans approximately $190 billion in capital expenditures for calendar 2026, including roughly $25 billion attributed solely to higher component prices. This weighs heavily on free cash flow and represents an enormous wager on sustained AI demand. On the analyst front, JPMorgan’s Samik Chatterjee raised his 2027 price target to $625 from $550 on August 13, projecting $41 billion in annual revenue potential from Microsoft 365 Copilot alone. Whether this vision materializes will become clearer on October 27, when Microsoft reports its first-quarter fiscal 2027 results.
In the interim, Microsoft has managed ancillary developments with minimal market disruption. Two weeks ago, the company shipped security updates patching 421 vulnerabilities, including a zero-day flaw in the Windows network driver that was already being actively exploited. The stock gained 3.0 percent in the following period, suggesting security incidents are now treated as routine operational matters rather than systemic risks. Additional initiatives—the consolidation of the Copilot app, membership in the Agentic AI Foundation with over 250 members, and strategic partnerships in the Philippines and Saudi Arabia—reinforce the platform strategy without independently moving the needle.
The underlying tension is structural. Microsoft is undeniably profiting from the AI boom; every new partnership, from Moonshot to Meta to Saudi Arabia, confirms this trajectory. However, a business model increasingly reliant on rented infrastructure rather than owned software may not deliver the same dreamlike returns as the licensing era. The market has decided the answer is yes, at least for now. The upcoming October earnings report will provide the first rigorous test of whether that conviction is justified.