Amazon Web Services—the profit engine that long powered Amazon’s ambitions—just hit a speed bump that sent its parent company’s stock sliding as much as 8% in after-hours trading. For the second quarter ending June 30, 2025, AWS delivered $30.9 billion in revenue, a 17.5% year-over-year increase. That would normally be a solid quarter, but placed next to Microsoft Azure’s 39% surge and Google Cloud’s 32% leap, it looked like a warning sign. The cloud market’s center of gravity has shifted, and investors are now punishing any provider that fails to capture AI-driven hypergrowth.

The Quarter That Spooked Wall Street

Amazon’s Q2 2025 numbers, released July 31, revealed a deepening strategic chasm. AWS’s operating margin contracted to roughly 32.9%, down from the high-30s range a year earlier, as the company ramped spending on AI infrastructure. While Amazon’s overall retail and advertising businesses performed well, analysts fixated on the cloud division because AWS generates a disproportionate share of Amazon’s operating income—$9.9 billion of the company’s total $14.7 billion in operating profit this quarter.

Microsoft’s Azure, by contrast, not only grew faster but did so with an AI narrative that resonated. The company disclosed that AI services contributed 8 percentage points to Azure’s growth, up from 7 points the prior quarter. Google Cloud, powered by its Gemini models and TPU-driven infrastructure, posted its strongest growth in two years, crossing $12 billion in quarterly revenue. CEO Sundar Pichai told analysts that AI solutions were “driving deeper, more valuable customer relationships.”

The after-hours reaction was swift and punishing. Amazon shares fell between 3% and 8% depending on the trading window, reflecting not just disappointment but fear that AWS—the original cloud giant—is losing its edge in the most important technology transition in a decade.

Why Investors Are Rewriting the Cloud Playbook

Wall Street no longer views cloud providers solely through the lens of compute and storage capacity. Today’s premium valuations are earned by:

  • Integrated AI products that bundle models, tooling, and applications into high-margin, sticky subscriptions.
  • Custom silicon that reduces dependence on third-party GPUs and drives down training costs.
  • Ecosystem lock-in through AI features woven into productivity suites, collaboration tools, and search interfaces.

Microsoft and Google have mastered this formula. Copilot now permeates Microsoft 365, Azure AI Studio, and GitHub, turning every enterprise license into a potential AI upsell. Google has embedded Gemini into Workspace, cloud databases, and its advertising stack. In both cases, AI isn’t a separate line item—it’s a multiplier on existing cloud contracts.

AWS, meanwhile, still relies heavily on its traditional infrastructure-as-a-service model. Amazon CEO Andy Jassy insisted on the earnings call that AI adoption is “still early” and that AWS will benefit as companies move from experimentation to production. But investors clearly want to see faster translation of AWS’s vast resources into AI-native revenue.

AWS’s AI Arsenal: Bedrock, Trainium, and the Anthropic Bet

AWS isn’t standing still. The company has built one of the most comprehensive AI infrastructure stacks in the industry, anchored by:

  • Amazon Bedrock, a managed service that gives customers access to models from Anthropic, Meta, Mistral, and others through a single API.
  • Custom silicon, including the Trainium2 training accelerator and Inferentia2 inference chip, both designed to lower the cost of large-scale AI workloads.
  • Strategic partnerships, most notably an $8 billion investment in Anthropic, the maker of the Claude model family. That deal designated AWS as Anthropic’s primary cloud and training partner, a move aimed at locking in the AI lab’s massive compute demands for years.

Financial analysts at firms like UBS and Bernstein have begun modeling scenarios where Anthropic alone could contribute up to $5–$10 billion in annualized AWS revenue by 2027—if the partnership deepens as planned. But that future revenue hasn’t materialized fast enough to satisfy investors watching Azure and Google Cloud pull away in the present.

The CAPEX Dilemma: Spending Hundreds of Billions with No Guarantee

Amazon has telegraphed a colossal capital spending plan for 2025, with company statements and CNBC reports pegging the figure at over $100 billion annually, predominantly for AI and cloud infrastructure. Some aggregated industry projections push the number as high as $118 billion when multi-year commitments are included, but the exact figure depends on accounting treatments and lease structures.

This spending spree is necessary but painful. Every new data center, every Trainium wafer, and every megawatt of power capacity pressures near-term margins. Jassy told investors the company is “managing through supply constraints” for advanced components and that capacity should improve in the second half of 2025, but he offered no firm timeline for when the spending might translate into reaccelerating growth.

Microsoft and Google face similar constraints. Both acknowledged that AI demand exceeded capacity in certain regions, and both are spending aggressively—Azure’s CAPEX alone surged 78% year-over-year. The difference is that those companies can point to immediate, AI-driven revenue boosts that justify the outlay. AWS cannot yet make the same claim with the same force.

The Competitive Landscape: Integration vs. Modularity

Azure’s secret weapon is integration. By embedding Copilot into Word, Excel, Power BI, and Teams, Microsoft has made AI consumption an organic extension of everyday work. That creates a flywheel: every Office 365 seat becomes a potential Azure AI customer, and every AI feature deepens the company’s moat.

Google Cloud’s playbook is similar but distinct. Gemini works across Gmail, Docs, Sheets, and Google Meet, while Vertex AI provides a unified platform for building and deploying custom models. Google’s search and advertising businesses also provide a near-infinite reservoir of AI training data and monetization paths that AWS doesn’t have.

AWS’s modular approach—APIs, building blocks, high configurability—appeals to developers and regulated industries that demand flexibility. But it’s slower to convert into the kind of turnkey enterprise adoption that CFOs like to buy. As one infrastructure analyst noted, “AWS sells Lego bricks; Microsoft and Google are selling fully assembled houses.”

Leadership Speculation and Corporate Rumors

The growth gap has ignited chatter about whether AWS needs a leadership shakeup. A handful of investor threads and tech blogs speculated about a possible return of Amazon founder Jeff Bezos to a more active role. No credible corporate filing or major news outlet has substantiated those rumors. They appear to be exactly what they sound like: the market’s instinct to personalize a structural challenge.

What’s more concrete is the pressure on the current leadership team. During the Q2 call, analysts repeatedly pressed Jassy and AWS CEO Adam Selipsky on the timeline for AI revenue acceleration. The responses were measured but defensible: AWS has the broadest customer base, the deepest partner ecosystem, and an unmatched track record of turning infrastructure investments into durable franchises. The question is whether the organization can move fast enough to match the AI-centric growth rates of its rivals.

Opportunities That Could Rewrite the Narrative

Three catalysts could shift sentiment in AWS’s favor within the next 12 to 24 months:

  1. Anthropic monetization at scale. If Claude and other Anthropic models become the go-to choice for enterprises building generative AI applications, the AWS tie-up could generate billions in high-margin computing and model-hosting revenue. Early indicators are promising—Anthropic’s valuation hit $61.5 billion in its latest funding round, and its compute needs are growing rapidly.
  2. Trainium and Inferentia adoption. Should AWS convince large customers like Salesforce, Intuit, or even net-new AI startups to train on Trainium rather than Nvidia GPUs, it could lower its own costs and create a pricing umbrella that competitors can’t match. The first production-scale deployments are underway, but mass adoption remains a future milestone.
  3. Bedrock as an enterprise AI gateway. If AWS can turn Bedrock into a one-stop shop for governed, secure AI—with built-in retrieval-augmented generation (RAG), agents, and fine-tuning—it could close the productization gap with Microsoft and Google. The service has grown quickly since its 2023 launch, but it’s still viewed primarily as a model catalog rather than a full-stack AI platform.

Checklist for Observers: What to Watch Next

For investors, enterprise IT buyers, and cloud watchers, these metrics will determine whether AWS regains momentum:

  • AI-revenue disclosure: Will AWS break out AI-specific revenue in future quarters? That would signal confidence and give analysts a clear metric to track.
  • Margin trajectory: Operating margins need to stabilize above 35% once current CAPEX cycles conclude; anything lower suggests the investments aren’t generating sufficient returns.
  • Anthropic consumption: Public references to Anthropic’s AWS spending—via earnings calls or securities filings—will be a leading indicator of the partnership’s financial impact.
  • Trainium readiness: As Trainium2 becomes generally available, the number of large-scale training clusters running on AWS silicon will be a key competitive signal.
  • Product launch cadence: The frequency and depth of Bedrock updates, new AI services, and integrations with existing AWS tools will show whether the company is shifting from infrastructure to solutions.

The Bottom Line: Realignment, Not Collapse

AWS remains the largest cloud provider by revenue—its $30.9 billion quarter dwarfs Azure’s roughly $29 billion and Google Cloud’s $12 billion. It still generates healthier margins than most competitors and benefits from a vast, loyal enterprise customer base. Writing it off would be a mistake.

But the market has changed, and the rules of leadership have changed with it. Growth is no longer a function of capacity; it’s a function of AI productization. AWS has the assets—custom silicon, a deep partnership roster, and a soon-to-be-released next-gen model marketplace—to compete. What it doesn’t yet have is proof that it can turn those assets into the same kind of explosive, AI-fueled expansion that Azure and Google Cloud are already delivering.

Jassy’s “still early” refrain is factually accurate—enterprise AI adoption is in its first innings. But the market is already pricing in the later innings, and AWS needs to start putting runs on the board now. If it can do that, the current stock slump may look, in hindsight, like a buying opportunity. If it can’t, the cloud market’s historic leader risks becoming a fast follower in the AI era.