And Microsoft reported their earnings, and a lot of interesting details came from that.
Microsoft’s capital expenditures reached a record high of $37.5 billion in the latest quarter, a significant increase of nearly 66% from the previous year. Despite this surge in spending, the company’s revenue growth only slightly exceeded Wall Street expectations, rising 17% to $81.3 billion, which was above analysts’ predictions of $80.27 billion. The Azure cloud division reported a 39% growth, just surpassing the expected 38.8%. However, concerns linger about Microsoft’s reliance on OpenAI, which significantly influences its cloud backlog. Approximately 45% of Microsoft’s remaining performance obligations are tied to OpenAI.
Despite overall growth, the Windows devices segment saw only a 1% increase, while Xbox content and services declined by 5%.
Microsoft has begun utilizing its latest in-house AI chip, named Maia 200, as part of its strategy to reduce dependence on Nvidia’s advanced chips. According to Scott Guthrie, executive vice president at Microsoft, these chips are currently operational in one data center, with plans for global rollout. The adoption of these chips is expected to lower the costs associated with running AI applications, including the Microsoft 365 Copilot, which leverages AI models from OpenAI and Anthropic for task automation in Office 365. It remains uncertain if servers equipped with Maia 200 will be made available to Microsoft Azure customers, following the company’s recent adjustments to its rollout plans due to previous delays.
Why do we care?
Microsoft just told you, very clearly, that AI is no longer an add-on. It’s the load-bearing wall. And when something becomes structural, the company stops optimizing for flexibility and starts optimizing for control.
That $37.5 billion isn’t speculative. It’s already spent. Which means the question isn’t if Microsoft will recoup it—it’s how. History suggests that recouping that investment will come through pricing power, architectural lock-in, and partner margin compression—not shared upside.
The OpenAI dependency is doing two things at once. It’s driving massive Azure demand, and it’s forcing Microsoft to build escape hatches. Maia is one of those escape hatches—not for customers or partners, but to rebalance Microsoft’s own dependency risk.
If you’re an MSP telling customers to “just trust Copilot” or “build deeper into Azure AI” without talking about cost drift, authority, or exit strategy, you’re not being strategic. You’re outsourcing judgment to a vendor that just signaled it will prioritize internal economics first.
The real risk isn’t downtime. It’s automation authority without accountability. When AI systems act, who owns the decision? Microsoft is quietly answering that question in its favor.
This matters now because the next wave of changes won’t be framed as breaking changes. They’ll be framed as “enhancements,” “optimizations,” or “alignment.” And by the time customers notice margin pressure or governance gaps, the architecture will already be locked in.
That’s the lesson here: Microsoft is building permanence. MSPs need to build optionality—or accept that they’re being positioned as risk absorbers in someone else’s platform strategy.