1. Amazon AI business run rate passes $25 billion: what was reported?
Amazon said in its second-quarter 2026 results, released on July 30, that its AWS artificial intelligence business and its custom chip business have each passed an annualized revenue run rate of $25 billion, with both growing at triple-digit percentages year over year. The Amazon AI business run rate arrived alongside AWS revenue of $42.2 billion, up 37%, and a capital spending forecast of about $220 billion for 2026.
The disclosure matters because it gives investors a number for a business that had been described mostly in terms of demand and spending. It also lands in the middle of a debate over whether the enormous sums that large technology companies are committing to AI infrastructure will be repaid by customers willing to pay for the capacity.
2. What does a $25 billion run rate actually measure?
A run rate takes the revenue earned in a recent period and multiplies it to cover a full year. If a business earns a certain amount in a month or a quarter, the run rate shows what it would earn over twelve months if that pace simply continued. It is a snapshot of momentum, not a figure of revenue already booked across a calendar year.
That distinction is important for fast-growing units. A division expanding at triple-digit rates will usually book less in a trailing twelve months than its current run rate implies, because earlier periods were smaller. The figures are also reported by the company and are not audited as separate line items, so they are best read as management's own description of scale.
Two separate businesses crossed the threshold. One is the AI services sold through AWS; the other is the custom chip operation, which designs processors that customers rent in the cloud. Each is reported as exceeding $25 billion annualized, which means the combined AI-related activity is larger than either number alone, although the company did not present a single combined total in the figures cited here.
For readers new to the metric, the practical advice is to compare run rates with reported revenue over several quarters. When the two converge, the run rate was a reliable guide; when they diverge, the growth may have been lumpy or driven by one-time usage. Because Amazon has not broken out the AI and chip lines in its audited segments, the disclosure is a directional signal and not a full financial statement.
3. How did the rest of AWS perform in the quarter?
AWS, the cloud division that houses these AI products, reported quarterly revenue of $42.2 billion, a 37% increase from a year earlier. Annualized, that quarter equals a run rate of $169 billion. The growth rate was the fastest in 18 quarters, a sign that the cloud business has been accelerating rather than maturing.
The division was also highly profitable. AWS operating income reached $16.6 billion, compared with $10.2 billion in the same quarter a year earlier, for an operating margin of 39.4%. Strong margins matter in this context because the AI build-out is expensive, and a profitable core business is what funds it.
Company-wide, total net sales were $200.6 billion, up 20%. Amazon's reported net income was inflated by non-operating gains of about $53.4 billion tied mainly to its investments in Anthropic, so the headline profit figure does not reflect day-to-day operating results.
4. Why is the backlog $496 billion, and why does capacity matter?
Amazon reported an AWS backlog of $496 billion. Backlog counts contracted revenue that customers have committed to but that has not yet been recognized, which makes it a rough indicator of future demand. A figure of this size, equal to roughly 2.5 times the level a year earlier, suggests customers are signing long-term commitments for cloud and AI capacity.
The constraint is supply. CEO Andy Jassy said Amazon will not have enough capacity to meet all of its 2026 demand, and he expects the same dynamic in 2027. That is an unusual position for a retailer-turned-cloud-provider: orders exist that the company cannot yet serve. Revenue cannot be recognized until servers, chips and data center space are in place, so the speed of construction directly affects how quickly the backlog turns into reported sales.
Backlog is not the same as cash. Contracts can be renegotiated, timelines can shift and customers can consume services more slowly than planned. Even so, a commitment pool of this scale gives a cloud provider unusual visibility into the coming years, which is why analysts follow the number closely each quarter alongside revenue and margin.
The capacity limit also has a competitive dimension. Customers that cannot get the computing power they need from one provider may look to another, so a shortage can protect margins in the short run while creating pressure to build faster. The company has framed the situation as one of strong demand rather than weak execution, and the quarter's growth rate lends some support to that framing.
5. What is behind the $220 billion capital spending plan?
Amazon raised its 2026 capital expenditure guidance to about $220 billion, up from roughly $200 billion. The increase was attributed mainly to higher memory costs, and most of the spending is aimed at AWS and generative AI infrastructure. In the second quarter alone, cash capital expenditures were $53.1 billion.
Rising component prices have a direct effect on returns. If memory becomes more expensive, each dollar of spending buys less capacity, which can slow the conversion of backlog into revenue even when demand is strong. Investors therefore watch not only the total budget but also how efficiently it turns into usable computing power.
Large capital programs of this kind are usually spread over several years because data centers, power connections and specialized hardware take time to deliver. The 2026 figure therefore reflects commitments that will generate revenue in later periods, and the gap between spending today and income tomorrow is the source of the cash-flow pressure described below.
6. What does the spending do to Amazon's cash flow?
The cost shows up in free cash flow, which is the cash left after operating needs and investment in property and equipment. For the trailing twelve months, Amazon reported free cash flow as an outflow of $7.6 billion. The company linked the decline mainly to a $66.1 billion year-over-year increase in purchases of property and equipment, driven primarily by investment in artificial intelligence.
That leaves shareholders with a trade-off. One side of the ledger shows the fastest AWS growth in 18 quarters, a 39.4% operating margin and a very large backlog. The other shows negative trailing free cash flow and a spending plan that grows rather than shrinks. Whether the balance is favorable depends on how quickly the new capacity is filled by paying customers.
Free cash flow is a stricter measure than accounting profit, which explains how a company can post very large net income and still show an outflow. The profit figure in this quarter included gains on investments that did not involve cash from operations, while the cash-flow figure captures money actually spent on equipment and facilities. Reading the two together gives a more complete picture than either does alone.
7. How does Amazon's custom chip business fit in?
Designing its own processors lets a cloud provider offer customers an alternative to chips from outside suppliers, and it gives the provider more control over cost and supply. Amazon's chip business passing $25 billion annualized, with triple-digit growth, shows that the strategy has moved beyond an experiment and into a significant revenue stream.
It also helps explain why the company is willing to spend so heavily. Capacity that Amazon builds around its own silicon serves two purposes at once: it supports its AI services and it creates a product that customers can buy directly. The reported growth rate indicates that demand for that product is strong at present.
8. Why does this matter beyond Amazon?
Amazon is one of the largest buyers of AI hardware and one of the largest sellers of AI computing, so its numbers are watched as a gauge of the wider industry. A $25 billion run rate in AI services shows that customers are paying for the technology at scale, at least within the largest cloud platform. The spending plan shows how much capital the sector believes it needs to keep pace.
The results also illustrate the dependencies in the AI supply chain. Memory prices, chip availability and power capacity all feed into how much computing a dollar of investment can deliver. A change in any of them can shift the economics for every large provider at the same time, which is why a cost increase flagged by one company tends to draw attention across the market.
9. What should investors watch next?
The next indicators are the ones the results themselves highlight. AWS growth will show whether the acceleration of the second quarter persists. Backlog conversion will show how fast contracted demand becomes recognized revenue as new capacity comes online. Operating margin will reveal whether higher memory and infrastructure costs begin to compress profitability.
Free cash flow is the fourth measure. A recovery from the $7.6 billion outflow would suggest that the heaviest phase of investment is starting to pay off, while a deeper outflow would point to continued pressure. Amazon's next quarterly report will provide the first update on all four, and Jassy's comments on capacity for 2026 and 2027 set the expectation that supply, rather than demand, remains the main limit on growth.