Amazon’s invisible AI edge

Wall Street missed the signal. Claude, chips, and cloud are telling a different story

Amazon reported earnings this week and beat expectations across revenue, earnings, and operating income. And yet, the stock ended the week down more than 8 percent.

That reaction surprised me. Not because AWS growth lagged Azure and Google Cloud, that part is well understood, but because the market seems to have misread what Amazon is building underneath: a vertically integrated AI stack that is locking in enterprise use cases and long-term margin leverage.

This post is an attempt to surface that story and highlight what may be one of the most strategically important CapEx cycles in tech right now.

Claude and capex: the real Q2 story

Amazon spent $31.4 billion in capital expenditures during the quarter. That is more than Google or Meta. This spending is not about headlines. It is about infrastructure: custom AI chips, massive compute clusters, and a new generation of enterprise-ready AI platforms.

At the center of this buildout is Claude. It is the fastest-growing model on AWS and the focal point of Amazon’s multibillion-dollar partnership with Anthropic.

Claude is not just performing well in benchmarks. It is becoming embedded in real enterprise systems. Pfizer uses it to optimize mRNA yields, improving the efficiency of therapeutic production. Genomics England applies it to high-stakes genomic workflows. Tools like Sourcegraph and Cursor rely on Claude to write production-grade code. Every Claude model is trained and deployed on AWS. That includes Amazon’s Trainium chips, its Bedrock AI platform, and its developer tooling.

Claude is not just another model. It is a strategic layer Amazon now owns across training, inference, and enterprise integration.

AWS: still the profit core

AWS generated $30.9 billion in revenue, up 17.5 percent year over year. It delivered $10.2 billion in operating profit at a 32.9 percent margin. That accounted for more than half of Amazon’s total operating income for the quarter.

Those results were solid. But investors focused on the growth rate, which trailed Microsoft Azure at 34 percent and Google Cloud at 32 percent. What this misses is the root cause. AWS growth is not slowing due to demand. It is constrained by infrastructure and power availability. Management said these constraints will ease steadily over the next few quarters.

Meanwhile, demand is still rising. AWS’s backlog rose to $195 billion, up 25 percent from last year.

Anthropic: strategic integration, not just access

Amazon’s stake in Anthropic is not just about model access. It is about strategic alignment.

Claude 3.5 and Claude 4 are now widely considered key models for enterprise-grade coding. Adoption is growing fast in sectors like biotech, finance, law, and manufacturing. These are fields where small performance gaps lead to large productivity or cost gains.

Every Claude training run, every inference, every deployment happens on AWS. That includes Bedrock integration, hardware-level acceleration, and increasingly, workflow-level orchestration. This gives Amazon a structural position in the most valuable layers of enterprise AI adoption.

AWS is building the substrate of Enterprise AI

While Microsoft has aligned closely with OpenAI, and Google builds Gemini through DeepMind, Amazon is pursuing a different strategy. It is building the full substrate — the foundational infrastructure that makes enterprise AI usable, secure, and scalable.

Trainium2 chips now deliver 30 to 40 percent better price-performance than traditional GPUs. Amazon is building one of the largest AI superclusters on Earth to support future Claude training.

On the software side, Amazon has turned Bedrock into more than a model-hosting service. With AgentCore, it now provides enterprises with identity management, memory, and observability for deploying AI agents. This shifts AWS from model provider to automation platform.

What they’re building isn’t a prototype. It’s the backbone

Retail, Ads, and logistics are quietly thriving

Outside of AWS, Amazon’s other core businesses are showing strong momentum.

Online store revenue rose 11 percent year over year to $61.5 billion, an acceleration from 6 percent growth in Q1. Advertising revenue surged 23 percent to $15.7 billion, making it one of the fastest-growing ad businesses in tech. The international segment saw a 16 percent revenue increase and a 320 basis point expansion in operating margin.

Prime Day set new records for items sold and Prime sign-ups. Fulfillment speed, delivery density, and per-unit costs all improved. Amazon’s logistics network is becoming more efficient at scale, not less.

These segments are executing well, even if they are not driving the AI narrative.

CapEx isn’t cost. It’s leverage.

Amazon’s capital expenditures are projected to reach $118 billion this year. This level of spending is not just aggressive. It is aligned with a long-term infrastructure thesis.

This investment includes custom silicon, including Trainium and Inferentia. It includes global data center buildouts, superclusters, and dedicated power provisioning. It includes software platforms like Bedrock and AgentCore, which let enterprises deploy secure AI agents at scale. It also includes new developer tools, like Strands and Kiro, that simplify the creation of AI applications.

What sets Amazon apart is that it owns the stack. It is not just building infrastructure. It is monetizing every layer through AWS.

The competitive context

Microsoft is growing Azure quickly through its alignment with OpenAI. Google is gaining traction with its Gemini models and vertically integrated AI teams. Meta is spending heavily on open-source models and consumer LLMs.

Amazon’s position is different. It is not trying to win through model branding. It is building an enterprise AI platform it fully controls: from silicon, to model training, to deployment, to automation.

That structural leverage matters more as AI workloads move into production.

The Real Question

Much of the post-earnings conversation focused on AWS’s year-over-year growth rate.

But the more important question is not whether AWS can grow faster in the second half of the year.

It is whether any other company is building the kind of integrated, full-stack AI leverage Amazon is putting in place right now.

-Maureen