Amazon's AI Revenue Hits $15 Billion as Jassy Plans $200B Spend

Amazon just revealed a number it had never shared before — and it's enormous. In his annual shareholder letter released April 9, CEO Andy Jassy disclosed that AWS AI services have reached a $15 billion annual revenue run rate. That's not a forecast. That's what the business is generating right now.

To put that growth in context, Jassy noted that AI revenue at AWS is growing approximately 260 times faster than AWS itself did at the same stage of development. AWS took roughly 10 years to reach $15 billion in revenue. Its AI services got there in a fraction of that time.

Amazon's stock surged 5% on the announcement. Investors are clearly betting that AI is becoming the primary growth engine for the company's most profitable division.

The $200 Billion Question

The headline AI revenue number came alongside an equally staggering capital expenditure plan. Jassy confirmed Amazon expects to spend approximately $200 billion on capex in 2026, with the bulk directed toward AI infrastructure.

That figure raised eyebrows. $200 billion is roughly the GDP of Greece. It's more than what most countries spend on their entire defense budgets. And Jassy addressed the skepticism directly in his shareholder letter, arguing that the investment is grounded in customer demand, not speculation.

The spending breaks down across several categories: data center construction, networking equipment, and crucially, custom silicon. Amazon's custom chip business — including its Graviton processors, Trainium AI training chips, and Nitro security chips — has now exceeded a $20 billion annual revenue run rate, doubling from $10 billion earlier this year.

Custom Chips Change the Equation

Perhaps the most consequential detail in Jassy's letter was his statement that Amazon could begin selling its AI chips to external customers. That would put Amazon in direct competition with Nvidia and AMD in the AI accelerator market — a move that could reshape the semiconductor landscape.

Amazon's Trainium chips are already used internally for training large AI models. If those chips become available to third parties, it would give companies an alternative to Nvidia's dominant (and often scarce) GPUs. Given Amazon's scale in manufacturing and its ability to bundle chips with AWS cloud services, this could become a significant competitive threat.

The custom chip strategy also explains part of the massive capex. Building a semiconductor supply chain is expensive, but the margins are attractive once established. Amazon is essentially replicating the vertical integration playbook that Apple used with its M-series chips — designing silicon specifically optimized for its own workloads, then potentially licensing that advantage to others.

What's Driving the Demand

Jassy attributed the AI revenue surge to enterprise adoption accelerating faster than expected. AWS offers AI services across the full stack — from foundational model hosting through Amazon Bedrock to custom model training on Trainium to AI-powered application services.

The $15 billion figure captures revenue from all of these AI-specific services. It doesn't include the broader lift that AI workloads bring to general AWS compute, storage, and networking services — which means the total AI-driven revenue at AWS is likely substantially higher.

Enterprise customers are moving beyond experimentation. Companies that spent 2024 and 2025 running AI pilots are now deploying at production scale, driving the kind of sustained, high-volume usage that shows up as a $15 billion run rate.

Competitive Positioning

The disclosure positions Amazon as a serious contender in the AI infrastructure race, but the competitive landscape is fierce. Microsoft's Azure has the OpenAI partnership and a deeply integrated AI stack. Google Cloud has Gemini and a growing model ecosystem. Both have their own custom chip efforts.

What Amazon has is scale. AWS remains the largest cloud provider by revenue, and its breadth of services gives enterprise customers a reason to consolidate AI workloads alongside their existing infrastructure. The $200 billion capex commitment signals that Amazon intends to maintain that advantage through sheer investment volume.

Anthropic's presence on AWS through Amazon Bedrock is another strategic asset. With Anthropic's revenue run rate recently passing $30 billion, a significant portion of that demand flows through AWS infrastructure.

What This Means for the Market

Amazon's numbers confirm a broader trend: enterprise AI spending is not slowing down. The companies that control AI infrastructure — compute, chips, and cloud services — are capturing an outsized share of the value being created.

For investors, the $200 billion capex plan is a double-edged sword. It signals enormous opportunity, but it also means Amazon is spending heavily to capture it. The payoff depends on whether enterprise AI adoption continues accelerating or hits a plateau.

For enterprises evaluating cloud providers, the message is practical: AWS is all-in on AI, and the investment in custom chips and infrastructure should translate into competitive pricing and availability for AI workloads.

The AI infrastructure buildout is now the defining story in tech. Amazon's $200 billion bet makes that unmistakably clear.