AWS and Azure Deadlocked in $100K-$500K Cloud Spend, Google Chases

Roughly 40 percent of cloud customers spend between $100,001 and $500,000 per month on Amazon Web Services. For Microsoft Azure, the figure is 41 percent. The near-identical concentration, drawn from Flexera's new 2026 State of the Cloud report, captures the current equilibrium between the two largest hyperscalers more precisely than any market-share percentage could. The survey, which polled over 750 executive leaders and technology professionals during the winter of 2025, found the two providers deadlocked in the midrange enterprise spending tier that generates the bulk of recurring cloud revenue.

Published on April 8, the Flexera data provides the most granular public view of how real customers distribute their cloud budgets. Sixty-two percent of respondents were based in the Americas, 23 percent in Europe, and 13 percent in Asia-Pacific. The spending bands tell a layered story. At AWS, 9 percent of customers spend $500,001 to $1 million monthly, another 9 percent spend $1 million to $2 million, 6 percent spend $2 million to $5 million, and 5 percent exceed $5 million. Azure's high-end profile is nearly indistinguishable: 11 percent in the $500,001-to-$1-million band, 6 percent in the $1 million-to-$2 million range, 6 percent at $2 million to $5 million, and 5 percent above $5 million.

Google Cloud Platform tells a different story. GCP has the highest share of users in the lowest spend tier, with 20 percent reporting monthly bills below $50,000, according to the Flexera survey. Only 28 percent of GCP customers report spending between $50,001 and $500,000, and the drop-off above half a million dollars is steep: 6 percent spend $500,001 to $1 million, 5 percent spend $1 million to $2 million, and just 3 percent exceed $5 million. That distribution is the third quarter in a row in which Google Cloud's spending profile has skewed noticeably lower than its two larger rivals, a pattern that explains why the company's strategy has centred on moving existing customers up the spend curve through AI workload adoption.

The enterprise-only cut of the data sharpens the contrast further. Among enterprise respondents, AWS has its highest concentration in the $200,001 to $500,000 monthly range, where 16 percent of enterprises report spending at that level, ranking first among all providers. Azure leads in the $50,001-to-$100,000 and $100,001-to-$200,000 bands, with 15 percent of enterprises in each. Google Cloud's largest share of enterprise customers sits in the sub-$50,000 tier at 18 percent, ranking first among providers in that bracket. Oracle Cloud Infrastructure and IBM Cloud show even more modest profiles, with the majority of their enterprise customers clustered below $200,000 per month.

The Flexera numbers land at a moment when the hyperscaler pecking order is being rewritten by AI infrastructure demand. Worldwide cloud infrastructure service revenue reached $119 billion in the fourth quarter of 2025, up 30 percent year over year, according to Synergy Research Group data reported by CRN in February. For the full year, cloud service revenue surpassed $419 billion. John Dinsdale, chief analyst at Synergy, noted that growth rates of this magnitude had not been seen since early 2022, when the market was less than half its current size. Combined, AWS, Microsoft, and Google captured 68 percent of global cloud infrastructure spending in the fourth quarter.

Dinsdale told CRN that generative AI has "simply put the cloud market into overdrive." The statement is supported by forward-looking indicators that go well beyond quarterly revenue. By the first quarter of 2026, the three largest providers generated a combined $92 billion in cloud sales, with AWS reporting $37.6 billion, up 28 percent, its fastest pace in fifteen quarters. Microsoft's Azure grew roughly 40 percent in its fiscal third quarter, and Alphabet reported that Google Cloud grew 63 percent year over year to reach a $71 billion annual run rate.

The Capex Race Reshapes Product Strategy

Behind the revenue figures sits an infrastructure investment cycle without precedent. Amazon committed to $200 billion in capital expenditure for 2026, a number CEO Andy Jassy disclosed in his April shareholder letter alongside the news that AWS AI revenue had reached a $15 billion annualised run rate. The letter, analysed by 24/7 Wall St, framed the spending as a direct response to AI demand that Amazon believes is still in its early stages. Microsoft, not far behind, projected roughly $190 billion, while Google's Sundar Pichai told investors the company planned to spend upward of $185 billion.

What each hyperscaler is buying with that capital differs in ways that will shape product roadmaps for years. At AWS re:Invent in December 2025, CEO Matt Garman unveiled a suite of what the company calls "frontier agents," autonomous AI systems capable of executing multi-day projects without continuous human involvement. GeekWire reported that the announcements also included Trainium3, the latest generation of Amazon's custom AI silicon, and a new offering the company calls private "AI factories," dedicated infrastructure clusters for enterprise customers running large-scale training and inference workloads. The silicon strategy matters because it gives AWS a lever that neither Microsoft nor Google can replicate at the same scale: the ability to tune procurement economics across a vertically integrated stack from chip to cloud service.

Microsoft's product strategy in early 2026 has been defined by a multi-model approach. In April, the company released three in-house AI models, MAI-Transcribe-1, MAI-Voice-1, and MAI-Image-2, making them available through its Azure AI Foundry platform. Business Insider reported that the move signalled Microsoft reducing its reliance on OpenAI, its longtime partner, even as the two companies renegotiated their commercial arrangement. At the same time, Azure's 40 percent growth rate in the most recent quarter was fuelled in significant part by Copilot and agent-driven workloads that pull through compute, storage, and networking services the way database migrations once did.

Google Cloud's product story in the first half of 2026 revolves around Gemini 3 and the platform's persistent supply constraints. During Alphabet's fourth-quarter 2025 earnings call in February, Pichai told analysts that the company had been "supply constrained even as we've been ramping up our capacity" and expected to remain so through the year. Google Cloud reported a $240 billion revenue backlog, and Gemini Enterprise had already sold over 8 million paid seats. At Google Cloud Next in late April, the company announced a new platform for building and managing enterprise AI agents atop Vertex AI, extending Gemini's reach into automated business workflows. The strategy is straightforward: convert the supply-constrained demand into committed, multi-year contracts that move customers from the sub-$50,000 tier into higher spending bands.

Pricing Moves Signal Where the Margin Pressure Is

Amid the AI product barrage, the most revealing pricing signal of the past twelve months came from a moment at AWS re:Invent that had nothing to do with artificial intelligence. At the close of his keynote, Garman introduced Database Savings Plans offering up to 35 percent off on-demand pricing for Amazon Aurora, RDS, and DynamoDB. GeekWire's Todd Bishop reported that the announcement drew the loudest cheers of the keynote, a reminder that even in an AI-obsessed market, cost relief on core services remains the fastest way to an enterprise buyer's wallet. The Database Savings Plans are structured similarly to AWS's existing Compute Savings Plans, requiring a one- or three-year commitment in exchange for predictable discounts, a mechanism that locks in customer spend while lowering the per-unit price.

Microsoft is pursuing a different pricing transformation, one tied to the architecture of AI itself. During the company's fiscal third-quarter 2026 earnings call, CEO Satya Nadella outlined what CRN described as a shift to "per-seat plus consumption-based AI and agent pricing models." The model breaks from the traditional SaaS per-user license by layering a metered consumption charge on top of the base subscription whenever an AI agent executes a task. For customers, it promises alignment between cost and value. For Microsoft, it creates a revenue stream that scales with usage rather than headcount, a meaningful distinction as enterprises deploy agents that do the work of dozens of employees.

Google Cloud has leaned on a simpler lever: being the most cost-effective option among the big three for standard compute and storage workloads. Reports published on MSN in late April 2026 highlighted that Google Cloud had gained traction with the most cost-effective pricing among major providers, even as AWS prepared for its record $200 billion capex year. The Flexera data supports the competitive dynamic. GCP's concentration in the sub-$50,000 tier is both a vulnerability and an opportunity. It means Google Cloud is winning a large number of smaller accounts that its rivals are not, but it also means a smaller base of customers generating more than $1 million per month. The company's response has been to bundle AI services, most visibly Gemini for Workspace and the Vertex AI agent platform, in ways that raise the average deal size.

One structural shift worth tracking is Oracle's deepening relationship with the hyperscalers. In April, Oracle announced a private, high-speed interconnect with AWS, enabling seamless data movement between Oracle Cloud Infrastructure and AWS data centres beginning in US East. At Google Cloud Next, Oracle expanded its Google Cloud partnership, unveiling the Oracle AI Database Agent for Gemini Enterprise. The interconnects matter because they reduce egress costs and latency, the two practical obstacles that have historically prevented enterprises from running hybrid architectures across providers. For the hyperscalers, hosting Oracle workloads directly is a way to capture spend that would otherwise sit entirely outside their ecosystems.

The Flexera data also surfaces a FinOps reality that is easy to miss beneath the AI headlines. Across all providers, a significant slice of customers, 17 percent at AWS, 18 percent at Azure, and 20 percent at GCP, report spending less than $50,000 per month. These are not the AI training workloads that dominate earnings calls. They are the small and midsize businesses, the departmental deployments, the development and test environments that together form a long-tail revenue base. The question for each hyperscaler is whether AI tooling can pull those customers into higher spending tiers without triggering sticker shock that sends them to a cheaper rival.

Multi-cloud adoption, long talked about as an enterprise ideal, shows up in the Flexera data as a practical reality. The median enterprise respondent now uses three cloud providers, and the spending distributions across AWS, Azure, and GCP are correlated rather than mutually exclusive. A customer spending $500,000 per month on AWS is likely spending another $200,000 on Azure and $50,000 on GCP, not choosing one over the others. That pattern makes the pricing moves from each hyperscaler interdependent in ways that a single-provider analysis misses. When AWS cuts database pricing by 35 percent, it pressures not just Azure SQL and Cloud SQL but also the FinOps calculations that determine where the next workload gets placed.

For the remainder of 2026, the indicator to watch is whether Google Cloud can convert its supply-constrained backlog, now $240 billion, into spending that moves its enterprise customers out of the sub-$50,000 tier. The Flexera report will update again in twelve months, and the distribution of GCP spending across bands will be the cheapest signal of whether the strategy is working. The second indicator is how far Microsoft's shift to agent-based consumption pricing travels beyond Azure AI workloads and into the broader Microsoft 365 and Dynamics installed base, a move that would pull forward revenue from seat-based licensing into metered cloud consumption. AWS, with its Database Savings Plans and Trainium economics, is playing the longer, quieter game of making it increasingly expensive for a large enterprise to operate anywhere else. The Flexera numbers suggest it is working. The question is for how long.