Amazon Web Services posted a 28% year-on-year rise in net sales to $37.6bn in the first quarter, the company said on Wednesday. Andy Jassy, Amazon president and CEO, said the growth was the fastest for AWS in 15 quarters and tied it to the platform’s role supplying compute for generative AI. At the same time, Amazon warned it will keep increasing capital expenditure to build data centres, buy chips and kit, and expand power and networking. The scale of the spending will weigh on near-term cash flows even as Amazon frames the outlays as investments in long-lived assets.

Amazon’s cloud arm is growing at a pace not seen since early in its life. AWS net sales rose 28% from a year earlier to $37.6bn in the quarter. That rate, according to Andy Jassy, is the fastest in 15 quarters and reflects AI-related demand for large-scale compute.

Jassy put the surge in plain terms. "It’s very unusual for business to grow this fast on a base this large," he said, naming AWS’s role in the AI market. He noted that the company is seeing revenue from AI at a run rate the company never saw in its early years.

Growth tied to AI compute demand

AWS’s growth comes as firms race to deploy generative AI. Cloud providers sell the compute, storage and networking those models need. Jassy described AWS as a supplier of that compute. He blamed AI for a big chunk of the acceleration.

He drew a comparison to the platform’s early days. Three years after AWS launched it had a revenue run rate of about $58m, Jassy said.

In the first three years of the current AI wave, he added, AWS’s AI revenue run rate topped $15bn, roughly 260 times larger than that early figure.

Those figures show a shift in what customers buy. Where earlier cloud use centred on storage and standard compute, today’s projects need more specialised, high-power hardware. That includes GPU clusters and networking setups capable of moving vast amounts of data quickly.

Spending to build the stack

Growth is coming with a large spending bill. Jassy said AWS has to pay up front for land, power and buildings, along with chips, servers and networking gear. Those are necessary before customers pay for services.

Amazon described this as a timing mismatch. The company pays now. Revenue follows later. Jassy said those investments buy assets with long lives. Data centres are counted as assets lasting more than 30 years. Chips, servers and networking equipment are treated as having useful lives of five to six years.

Mastheads

Because Amazon is investing heavily in infrastructure, capital expenditure has been climbing. Jassy said capex growth will continue soon. He tied the pace of short-term spending directly to how fast AWS grows. "The faster AWS grows, the more short-term capex we’ll spend," he said.

Financial trade-offs and investor reaction

Investors often watch capex closely. Heavy investment can boost future revenue but it can also reduce free cash flow now. Jassy addressed that tension on the earnings call. He sought to reassure investors while acknowledging the immediate effect on cash generation.

He framed the current spending as a pay- now, benefit-later choice. Data centres and other infrastructure, he said, are long-lived assets that will support revenue for decades. Meanwhile, chips and servers will need replacement on a shorter cycle. The combination means Amazon is front-loading costs to meet demand today.

That pattern matters for Amazon’s financial profile. When capital spending rises faster than revenue, free cash flow tends to be weaker in the short term. Jassy warned that the company’s free cash flow will be affected while capex growth outpaces revenue.

Markets had already been pricing AI as a boon for the companies that sell compute. Amazon’s results add to that picture. AWS’s 28% leap is the latest data point showing how demand for AI infrastructure is routing revenue toward cloud suppliers.

How Amazon describes the investment horizon

Amazon treats its infrastructure outlays as investments rather than pure cost. That choice affects accounting and messaging. When the company buys land and builds a data centre, it records a long-lived asset on the balance sheet. The cost is depreciated over decades. When it buys chips and servers, the company depreciates those costs over a few years.

That accounting mix matters for cash flow timing and margin presentation. Large upfront capex depresses free cash flow in the short run. Depreciation spreads the expense over many years on the income statement. Investors and analysts parse both measures to judge how quickly the business will convert revenue into cash.

On the earnings call, Jassy emphasised the scale of the prize. He said Amazon is already a leader in the AI market and that companies continue to choose AWS for AI workloads. The implication is that the company is prioritising capacity to meet that demand even if it hits near-term cash metrics.

Building out capacity brings operational demands. Sites need power and cooling. They need secure networking and enough connectivity to customers. Amazon has signalled it will keep spending on those fronts. That includes land purchases and upgrades to electrical and networking infrastructure.

Those tasks are capital intensive. Land costs vary by region but can be sizeable near city centres and major hubs. Power infrastructure in many markets requires long lead times and large upfront investment. That combination explains why capex must be deployed before revenue follows.

At the same time, hardware lifecycles mean Amazon faces a steady refresh cycle. Chips and servers will be retired and replaced on a roughly five- to six-year schedule, according to the company. That creates recurring investment needs once the initial data centres are live.

Related Articles

"The faster AWS grows, the more short-term capex we’ll spend," said Andy Jassy, Amazon president and CEO.

This article was created with AI assistance.