Microsoft earnings: Turning up the heat with Azure, AI and the data center arms race – News

By all accounts, Microsoft Corp. just delivered one of the most powerful quarters in its history. But this wasn’t just a blowout print, it was a signal that Microsoft is playing a new game.

The company’s performance is redefining what hyperscale looks like in the artificial intelligence era. An objective analysis of the Q4 FY2025 earnings call suggests that Microsoft is firing on all cylinders in cloud, AI and infrastructure execution, and it’s building for a future measured in gigawatts, not racks.

The only potential negatives for investors are: 1) Capital expenditures are expected to be more than $30 billion in the coming quarter, signaling continued spending ahead of revenue and negatively affecting cash flow; and 2) Gross margin pressure continues in cloud from the lower margins realized in its Azure/infrastructure business.

Nonetheless, Microsoft continues to execute well with Azure at the center of its innovation engine. As such, our current estimates call for Azure revenue to approach $87 billion this calendar year, which represents a 37% increase in constant currency relative to CY 2024. This puts Azure’s growth rate on par with Google Cloud Platform, which we estimate is 1/3rd Azure revenue; and more than double the revenue growth rate of Amazon Web Services.

Azure: Still the engine, now with AI turbochargers

Microsoft reported that Azure surpassed $75 billion in annual revenue, up 39% year-over-year in constant currency. But this top line only tells part of the story. Our research indicates Azure’s growth is now fueled by a three-pronged engine: legacy migrations (e.g., Nestlé’s massive SAP shift), the rapid scaling of cloud-native workloads, and new AI-specific deployments. We estimate that Azure AI services contributed to 19% of Azure’s growth in the quarter, exceeding $3 billion.

Notably, Microsoft claims to have taken AI infrastructure share every quarter this year. Part of this can be attributed to the change that Microsoft made in its reporting. As we reported last year, Microsoft re-categorized its Azure definitions, removing some legacy businesses in decline and replacing those with AI services. We reported that this likely had the effect of lowering overall Azure revenue but jacking up its growth rate.

At the time we predicted that once Azure revenue became large enough to hide the fact that Azure revenue wasn’t as large as observers thought, Microsoft might share revenue guidance with Wall Street, which is exactly what happened in its latest print. This has allowed Microsoft to create momentum and the perception that it isn’t just keeping pace, rather it’s setting it.

A notable nugget for data platform watchers is with Azure Databricks and Snowflake on Azure both accelerating, Microsoft is not only winning on core infrastructure but also embedding itself deeper in the data layer that powers AI and filling gaps in its data offering with partners.

The Azure revenue trajectory has been quite remarkable as shown in the chart below. Entering COVID, we estimate that Microsoft Azure was generating roughly $4 billion per quarter. Today, our estimates indicate that Azure is running at more than $21 billion per quarter growing in the mid-to-high 30% range. Note the revenue flattening in 2024 roughly coincides with the removal of the Enterprise Mobility, Security and Power BI per-user pricing revenue. Then it steeply rises into 2025 as the Azure AI services kick in and Microsoft discloses that Azure has surpassed $75 billion.

Note: Microsoft’s FY 2025 ends in CY Q2 2025 and the $75 billion figure roughly corresponds to the CY periods comprising Q3 2024 through Q2 2025. In our view, Microsoft cleverly timed its reclassification of Azure revenue and the $75 billion disclosure to underscore its substantial revenue base and its significantly higher growth rate relative to AWS.

Update: Amazon Stock Down as Street Compares AWS and Azure Growth Rates

The narrative around Amazon is nuanced and possibly off base – i.e. specifically Amazon is down because AWS’ growth rate lags its competitors. Consider this: When AWS was the same size as GCP it was growing at 45% YoY, compared to theCUBE Research estimates that GCP is growing at around 40% currently.

As it pertains to Azure  we’d make two points: 1) Microsoft changed its reporting for Azure to eliminate low growth / declining security and some other line items and replaced it with AI, which lowered the base (not reported) and increased the growth (reported); and 2) Much of Azure’s “pop” is due to inference for ChatGPT. This is the AI component that Microsoft added into its Azure definitions.

If you take out the AI uplift from ChatGPT and normalize the Azure data, Azure’s growth would be around 20% and come in at around $19B last quarter. You have to go back to the 1st half of 2022, when AWS’ quarterly revenue was around $19B. AWS’ YoY growth at that time was a robust 36%.

So when you normalize for the ChatGPT inference lift, one could argue AWS is actually outperforming.

To add additional color to this analysis, George Gilbert points out that a big part of what’s driving the dramatic acceleration in Azure is inference payments they’re getting from OpenAI. Even for free ChatGPT users, OpenAI has to pay Microsoft. In fact, 50% of OpenAI revenue goes to COGS, most of which goes straight to Azure. ChatGPT weekly users accelerated from 300M in Q4 when they started rolling out new consumer features to 700M just days ago. GPT5, due any week, unfies speech, video, reasoning, tools, etc in ChatGPT. It’s going to be orders of magnitude more compute-intensive. We should see Azure growth accelerate even more, into the 40s next quarter and beyond. OpenAI is using so much compute that even their overflow vendors (Oracle, Stargate, Google) can’t satisfy all their demand. But Microsoft doesn’t want to publicize that much of their momentum comes from grabbing a ChatGPT tiger by the tail.

Rather, amazing results without attributing them to OpenAI cast a halo over all their products.

It’s a brilliant positioning and timing by Microsoft and weighing on Amazon shares. But observers should carefully watch the degree to which AI from ChatGPT is contributing to Microsoft’s Azure results.

The rise of the gigawatt cloud

There has been a lot of noise in the market around hyperscalers racing toward gigawatt-scale data centers. Microsoft cut through that noise with KPIs. According to the company, it stood up more than 2 gigawatts of new capacity in the last 12 months. The cloud giant now operates over 400 data centers across 70 regions, the largest global footprint of any cloud provider.

Microsoft claims that every Azure region is now AI-first, fully capable of supporting liquid cooling. This is not trivial. Liquid cooling enables Microsoft to increase the density of compute and improve asset fungibility, to balance global graphics processing unit supply and AI demand. In our view, this design shift is an underappreciated lever for long-term infrastructure margin improvement for Microsoft. In addition, the company’s posture forces AWS to explain the nuances of its availability zone strategy, which is often lost on customers that are not tech-savvy.

Key differences in data center footprints between Microsoft and AWS are articulated below:

Number and focus: AWS generally touts a larger number of regions and Availability Zones with a broader global reach, while Azure prioritizes locations with specific compliance needs and hybrid cloud solutions.

Approach to redundancy: Both platforms use Availability Zones to ensure high availability and fault tolerance. However, AWS’s multi-AZ design within each region is a core strength and ostensibly provides better uptime. Azure’s Availability Zones are also critical for redundancy, but are often augmented by region pairing for additional geo-redundancy.

Hybrid cloud strategy: Azure’s integration with the legacy on-prem Microsoft ecosystem gives it an advantage in hybrid cloud deployments with Azure Stack . AWS focuses on extending its cloud to on-premises environments with AWS Outposts, but it is a distinct approach and generally perceived as less mature than Azure Stack. Don’t confuse Azure Stack with Azure Arc. Azure Stack is a family of on-premises or edge products, while Azure Arc is a bridge that extends Azure management and services to any infrastructure, including Azure Stack Hub and Azure Stack HCI. Azure Stack HCI has been renamed to Azure Local.

Edge computing: Microsoft’s use of modular data centers for edge computing is a more pronounced aspect of their strategy compared to AWS’ approach, which, according to AWS, primarily relies on Local Zones and Wavelength for bringing services closer to users.

In essence, while both Microsoft and AWS offer robust cloud infrastructures, their strategic decisions regarding data center footprints reflect their target markets, product ecosystems and overall business models. AWS prioritizing cloud-native and Microsoft leveraging its vast legacy installed base to keep customers on its stack.

Capex explosion underscores the data center supercycle

Perhaps most notable in the earnings print is Microsoft’s ability to drive record capital expenditures while maintaining financial discipline. The company invested $24.2 billion in FYQ4 ending June 2025, including $6.5 billion in finance leases, and it’s guiding to more than $30 billion in Q1 FY26. More than half of this investment is in long-lived assets, with the rest going into GPUs, CPUs, networking and storage.

Despite its aggressive capex Chief Financial Officer Amy Hood was clear that this spend is tightly coupled to the company’s $368 billion commercial backlog. In other words, Microsoft isn’t betting on speculative demand, rather it’s building against committed revenue. That’s a level of alignment among sales, operations and infrastructure that few in the industry can match. Moreover, it underscores the undersupply of AI capacity that Microsoft currently has.

Software-led yield gains: The secret weapon

Nadella made a subtle but important point in our view: There’s a difference between being a hoster and a hyperscaler and that difference is largely seen in software. Nadella cited a 90% year-over-year improvement in token throughput per GPU for GPT-4o, purely through software optimization. This means Microsoft is squeezing more performance out of the same silicon, raising effective yield without raising its hardware costs. That’s a critical advantage in a world constrained by AI compute availability.

One could argue that this is also a shot across the bow at AWS. Nobody would argue that AWS is not a hyperscaler. However, though AWS is industry leading with regard to silicon design, hardware optimization and cost efficiency, its business model is essentially that of a hardware company with a distribution channel serving a massive ecosystem.

Microsoft, while perhaps not as proficient in hardware engineering, is capable in that regard and it also has a software-as-a-service business model that we estimate will approach $100 billion in CY 2025. This gives Microsoft a considerable margin advantage over any other cloud competitor. Despite wading so heavily into the cloud infrastructure business, Microsoft’s company-wide gross margin at 69% is substantially higher than that of Amazon (mid-to-high 40s company-wide) and Alphabet (high 50s company-wide). This is directly attributable to Microsoft’s contribution from software and its superior marginal economics.

Infrastructure margins under pressure, but manageable

As expected, Microsoft Cloud margins declined two points year-over-year, to 68%, and Azure’s segment margins were down four points. But management’s tone suggests this was not only anticipated but planned. The company claims it is intentionally scaling AI infrastructure ahead of revenue realization, a pattern we’ve seen before in prior platform shifts. With multiyear commitments on the books, we believe those margins will recover over time as utilization ramps and software-driven efficiencies compound.

Final takeaways

In our view, Microsoft’s Q4 wasn’t just a beat, it was a statement. The company is positioning Azure to dominate in the AI era by combining unmatched scale, an aggressive data center buildout strategy, a disciplined capex model and a software estate that shields it from market fluctuations and provides a simplified on-ramp from on-premises to the cloud. We believe this puts Microsoft in a commanding position to lead not only in cloud, but in the next decade of intelligent infrastructure.

This is all taking place in a backdrop of what we call a data center supercycle. Our estimates indicate that data center spend for power, cooling, servers, storage and networking will hit $500 billion in 2026. This spend is growing at a 16% 10-year compound annual growth rate, with the AI portion of that buildout growing at a CAGR in the mid-20s. Microsoft for its part is not showing any hesitation in attacking that opportunity and is placing massive bets that this cycle has fresh, young legs that will lead the next generation of technology for decades.

Image: theCUBE Research/Grok