Microsoft hits $4T as Azure's non-AI cloud business drives Q2 earnings beat

Summary

Microsoft hit $4 trillion in market cap last week. The earnings beat reveals a structural advantage that extends well beyond AI hype.

More than half of Azure's 33% revenue growth in Q1 came from non-AI services. In Q2, the company did not break out comparable numbers but signaled that "core infrastructure"—traditional cloud compute, storage, and networking—was the main driver of Azure's 39% growth.

This inverts the expected narrative. Microsoft positioned itself as the AI winner through GitHub Copilot (scaling from roughly $500M ARR to likely billions), an early OpenAI deal with sweeping revenue-share and IP rights, and CEO Satya Nadella as a public founder who owns the AI conversation. Yet the earnings data suggests the real money comes from something older: enterprises migrating from on-premises data centers to cloud infrastructure.

Companies are shifting from buying their own IT equipment to renting compute, storage, and networking from Microsoft. Some of this is AI-adjacent, with firms preparing infrastructure for future AI workloads or building their own models on top of GPUs. Much of it is not. It is hard drives, databases, CPUs, and Ethernet.

Microsoft's non-AI cloud services carry roughly 73% gross margins, compared to 30-40% for AI infrastructure. Once a data center is built, incremental CPU and storage sales are highly profitable. AI infrastructure requires constant capex to keep up with demand. GPU supply constraints may also play a role. Microsoft can provision traditional infrastructure reliably, while AI GPU capacity remains a bottleneck.

How Microsoft reports this creates ambiguity. A company fine-tuning an open-source model on H100s shows up as "core infrastructure," not Azure AI services. ChatGPT's infrastructure costs, which run on Azure, may also fall into the traditional bucket rather than token revenue. The split between AI and non-AI is messier than headlines suggest.

Amazon Web Services grew at 17.5% in Q2, disappointing investors and CEO Andy Jassy. Microsoft's cloud unit grew at 39%. The gap reflects more than execution. AWS lacks a consumer AI flagship like ChatGPT, now estimated to hit 1 billion weekly active users this year. Anthropic is hitched to AWS, but that infrastructure deal shows up as traditional cloud capex, not API token revenue. Microsoft captured the OpenAI relationship early and has a consumer AI product with billion-dollar-scale revenue.

AWS faces a deeper problem. Without a dominant AI consumer product feeding back into enterprise adoption, it must compete primarily on cloud infrastructure price and features. The hyperscalers have implicitly agreed not to wage price wars, which would destroy margins for everyone. AWS grows the traditional business steadily but lacks the leverage that Microsoft's bundled portfolio provides.

Microsoft's broader advantage is integration. Enterprises already use Excel, Teams, and Outlook. Azure offers not just compute but tight integration with Microsoft 365 and enterprise software. Google has Gemini and frontier research but weaker enterprise SaaS penetration. AWS has customer breadth but no equivalent to Microsoft's productivity suite.

Microsoft stock is up 40% since early April, pushing its forward P/E above 33x, richer than Amazon and well above Google's roughly 18x multiple. The valuation bet assumes sustained AI-driven growth. If non-AI infrastructure becomes the real driver, the narrative shifts. Non-AI cloud is a mature, competitive business. A 33x multiple hinges on whether AI revenue keeps accelerating and offsets slower growth in the traditional stack.

Layoffs signal the pressure. Microsoft has conducted more layoffs in 2025 than in 2022, 2023, and 2024 combined. Nadella appears to be squeezing margins and organizational efficiency even as the company performs well, either to fund AI R&D or to prepare for slowing growth elsewhere.

A UBS survey in July showed cloud customers improving sentiment on spending after muted attitudes early in the year, when tariff and recession concerns were high. Companies are moving forward with cloud migration. The framing is revealing: decades into the cloud era, enterprises still describe migration as a decision to "get on the cloud," as if it were the internet boom. This is both bullish for TAM expansion and suggestive of what is being bought—infrastructure, not edge-case AI services.