AWS Still Owns 30% of the Cloud. The Moat Is Holding — and Quietly Under Siege.

In 2024, one division inside Amazon did something the rest of the company spends decades trying to do: it printed money. AWS booked $107.6 billion in revenue and $39.8 billion in operating income — a 37% operating margin in a year when most of retail was scraping for pennies.¹⁸ AWS is roughly 17% of Amazon's revenue and consistently more than 60% of its operating income.⁸ The bookstore grew up to become a power utility. And like a utility, its dominance looks permanent — until you notice the rates two competitors are charging to dig a parallel set of pipes.

The official story is that the cloud is a settled three-horse race: AWS won, Azure is the enterprise runner-up, and Google Cloud is the perennial also-ran losing money. Two of those three claims are now wrong. The lead is real, but it has stopped being an equilibrium and started being a siege.

The 30-point lead, and why it isn't the whole story

Start with the number that anchors every cloud conversation. As of Q2 2025, Synergy Research data puts AWS at 30% of the global cloud infrastructure market, Azure at 20%, and Google Cloud at 13% — together about 63% of a roughly $99 billion quarterly market that grew 25% year over year.⁷ A 30-point lead over the field is enormous; in most industries it would be game over. The trap is reading a static snapshot as a stable one. A market growing 25% a year is not a market dividing a fixed pie — it's one where the challenger who grows faster than the leader gains ground every single quarter, no matter how far behind they start.

Look down the growth row. AWS grew 19% in 2024.¹ Azure grew 34% for its fiscal year.³ Google Cloud's Q4 2024 revenue was up 30%.⁵ The leader is growing at roughly half to two-thirds the rate of both challengers. That's the mechanism of the siege, stated arithmetically: when the laggards compound faster than the leader off a smaller base, the gap narrows even while everyone's revenue rises. AWS isn't shrinking. It's just being out-grown on two flanks at once.

How a head start became a moat — and why moats erode

AWS's moat is genuine, and it was built first. The popular timeline says AWS was 'founded in 2006,' but the real story is messier and more revealing: the first AWS service, Simple Queue Service, appeared in preview in late 2004, S3 launched publicly in March 2006, and EC2 — the engine that made on-demand computing a thing anyone could buy — only arrived in public beta that August.⁶ Amazon spent years selling rented computers before anyone serious thought rented computers were a business. That head start hardened into the deepest moat in the category: the most services, the most regions, the most engineers who already know the tools, and switching costs that compound with every workload a customer migrates in.

But a head-start moat protects against latecomers, not against incumbents with a different moat of their own. That's exactly what Azure and Google brought to the AI era. Microsoft didn't have to win developers from scratch — it already owned the enterprise through Microsoft 365 and Windows, and it bolted a privileged OpenAI relationship onto that distribution. The result: an enterprise buyer evaluating AI workloads finds the path of least resistance runs through the vendor whose software already sits on every desk. Azure's 34% growth and its move past $75 billion in annual revenue is what ecosystem lock-in looks like when it's pointed at a new workload.³ AWS's moat is breadth; Azure's is the customer relationship it already had.

Google Cloud is the third front, and it's the one most people still misread. The cliché is that it's a distant, money-losing also-ran. The money-losing part is simply no longer true. Google Cloud reached $43.1 billion in full-year 2024 revenue and posted $12.0 billion in a single quarter — Q4 2024 — up 30% year over year.⁵ Its operating income has swung sharply into positive territory — from a loss position as recently as 2022 to billions annually by 2024.⁹ The 'also-ran' is now a profitable business buying its way into the AI-infrastructure tier with custom silicon and a planned ~$75 billion of 2025 CapEx behind it.⁵ Different moat again: not breadth, not the desktop — raw AI compute and improving unit economics.

But isn't AWS just winning quietly?

The honest counter is that this 'siege' reads worse on a slide than it does on the income statement. AWS isn't merely defending — it's accelerating. Q4 2025 AWS sales rose 24% to $35.6 billion, operating income hit $12.5 billion in the quarter, and full-year 2025 AWS operating income climbed to $45.6 billion from $39.8 billion.² Growth re-accelerated, profits grew, and Amazon is pouring something near $100 billion of 2025 CapEx mostly back into AWS to defend exactly this position.¹ A skeptic could fairly say: a business throwing off $45 billion in operating income while re-accelerating isn't under siege — it's winning.

That's true, and it's the right place to be careful. The siege thesis isn't that AWS is losing — it's that the terms of the contest changed. For a decade, AWS won by out-scaling: more services, cheaper, sooner. Scale still works, which is why its margins are gorgeous and its profits are growing. But scale is a defensive weapon now. The growth differential — 19% against 34% and 30% — doesn't close because AWS is big and profitable. It closes only if AWS out-innovates on the one axis where its head start gives it no advantage: the AI workloads that Azure and Google purpose-built moats to capture. A leader can be richer every year and still be losing the future that the richness was supposed to buy.

Microsoft's own filings list Amazon, Google, Oracle, IBM, and open source as competitors in the same breath — a tacit admission that nobody in this fight believes the standings are settled.⁴ AWS still owns the toll road it built first, and it's collecting more toll than ever. But it's now defending a 30-point lead against two rivals who stopped trying to copy that road and built their own, each leading into the one destination that wasn't on AWS's original map. The lead is real. The equilibrium is gone. AWS spent fifteen years learning to out-scale everyone — and just discovered that the next decade will be won by whoever out-innovates, a contest where a head start counts for nothing.