AWS Aims for $1 Trillion as AI Spending Soars

What happened

Direct answer: Amazon chief executive Andy Jassy has told investors that Amazon Web Services could become a trillion-dollar annual revenue business within five to ten years, backed by about $220 billion of capital spending in 2026, most of it for AWS, and even more spending in 2027. AWS is already growing fast, up 39% year on year, but investors reacted cautiously, sending Amazon shares down about 6% in after-hours trading, because the bill for the AI infrastructure boom is arriving before all of the profits.

The announcement came on Amazon’s quarterly earnings call on Thursday 29 October 2026, reported by TechCrunch. It was not a new product launch, and there was no new chip, model or service to try. It was something rarer and, for anyone who runs a business on the cloud, arguably more important: a public statement of how big Amazon thinks the AI cloud market can get, how much it is willing to spend to win it, and how long it expects that spending to last.

Jassy put it in deliberately historic terms. “What we’re building at Amazon is unlike anything that’s ever been seen before. There has never been a business with a trillion dollars of annual revenue,” he said on the call. He then gave the numbers that are supposed to make that claim credible. AWS is, in his words, “a $184 billion annualized revenue run-rate business, and yet the opportunity remains extraordinary.” He forecast growth “north of 40 percent in each 2027 and in 2028” and said that, on that path, AWS could “be a trillion-dollar revenue business in the next five to 10 years.”

Those two numbers deserve a pause, because they are doing a lot of work. A run rate of $184 billion means that if the most recent quarter’s revenue were repeated for a full year, AWS would bring in $184 billion. It is not the same as trailing twelve-month revenue, and it is not a promise that the next four quarters will match the last one. But it gives a sense of scale: AWS alone is already larger, by revenue, than most of the world’s biggest companies. To reach $1 trillion from $184 billion in five to ten years would require sustained growth at a pace that very few businesses of this size have ever managed, which is exactly why Jassy paired the forecast with a spending plan of matching scale.

That spending plan is the second half of the story, and the half investors focused on. “We still expect 2026 cash capex of approximately $220 billion — the majority for AWS — and expect to spend meaningfully more in 2027, with a clear line of sight to attract its financial returns,” Jassy said. Capital expenditure, or capex, is money spent on long-lived assets: in this case, data centre buildings, power systems, cooling, networking equipment and the processors that run AI models. It is not the same as operating costs like salaries or electricity bills, although those rise too. It is the upfront cost of building capacity that Amazon hopes to rent out, in slices, for years to come.

The reaction tells you how the market reads that trade-off. TechCrunch reported that shares were down about 6% on Thursday in after-market trading, after analysts on the call repeatedly pressed Jassy to explain the spending, what signals would show the growth was really coming, and what profit margins to expect. That is not a panic, and it is not a verdict that Amazon is wrong. It is the normal tension of an infrastructure boom: the company building the roads is asking investors to pay for the roads now, on the argument that the traffic will arrive later, and investors want evidence that the traffic is not just passing through but paying tolls.

The numbers in one place

It is easy to lose the thread in a call full of percentages, so here are the figures TechCrunch reported, with what each one actually means.

A useful way to think about $220 billion is to compare it with what a small business spends on technology in a year, or even what a mid-sized company spends in a decade. It is roughly the entire annual revenue of many large corporations, spent in a single year on buildings and machines. Amazon can do that because AWS generates enormous cash flow and because the parent company can borrow and reinvest at a scale no smaller cloud provider can match. That is precisely the point Jassy was making to analysts: Amazon believes it can fund this plan from its own resources, and that the investment will, in his words, “forever change the size of Amazon” and “forever change our revenue.”

He also framed the spending as a response to a moment that will not wait. He called the AI boom a “once in a lifetime opportunity” and said Amazon was “in the very early days” of AI adoption by enterprises, adding that so far “the strategy is working.” That phrase matters. Much of the current AI spending debate is about whether businesses are actually using AI in ways that pay for themselves, or whether they are experimenting while vendors build capacity on speculation. Jassy’s answer is that enterprise adoption is early but real, and that the constraint is capacity: customers want to run their core systems and their AI workloads in AWS, and Amazon is adding capacity each year to meet that demand.

Why AWS is spending like this now

To understand the spending, it helps to understand what an AI workload does to a data centre. A traditional cloud workload, like hosting a website or running a database, needs a modest amount of computing power, spread fairly evenly over time. An AI workload, especially training a large model or serving millions of chatbot and agent requests, needs huge bursts of specialised computing, often on graphics processors or custom AI chips, with very fast networking between them and enormous amounts of power and cooling. You cannot simply squeeze that into an old data centre. You need new buildings, new power connections, new cooling systems, and new chips, and you need them in the places where customers and power are available.

That is why the spending is not just about buying more servers. It is about building an industrial system: land, power contracts, grid connections, water and cooling, fibre, and the specialised hardware that sits inside. Each of those has a lead time measured in years, not weeks. If Amazon waits until demand is proven beyond doubt, it will be too late to serve it, because a competitor that built earlier will have the capacity. If it builds too early, it owns expensive buildings with empty racks. The $220 billion figure is Amazon’s bet that building early is the smaller risk.

There is also a competitive reason. Microsoft has just reported another quarter of record-breaking revenue, driven largely by its cloud business, and has also spent heavily on AI infrastructure. TechCrunch noted that other cloud providers are spending big but also earning big. In a market where capacity itself is a selling point, because customers cannot run AI workloads without it, falling behind on building is falling behind on selling. Jassy’s line, “We’re going to invest to continue being the leader,” is a direct statement of that logic.

A third reason is less visible but important for small businesses: Amazon is not only building for AI model training, which is dominated by a handful of huge customers. It is building for inference, the everyday running of AI features inside ordinary business software, and for enterprises moving their core systems to the cloud alongside their AI work. Jassy specifically mentioned “enterprises wanting to run their core and AI workloads in AWS” and “new core and AI functionality.” That is the market that eventually reaches a small firm in Leeds or Bristol: not a frontier model trained on a supercomputer, but a customer service tool, a forecasting model, or a document summariser that runs on AWS infrastructure and is billed by the minute.

How this compares with Microsoft and the rest of the cloud market

Amazon did not make this announcement in a vacuum. In the same week, Microsoft reported $33.2 billion of Azure revenue in dollars for the first time, recorded $10 billion of revenue from OpenAI, and posted its tenth straight quarter of record-breaking revenue. Those figures, also reported by TechCrunch, show a cloud market in which the largest players are all growing quickly and all spending heavily at the same time.

The comparison is useful because it stops the AWS story from being read as a single-company bet. When Microsoft, Amazon and Google all increase capital spending in the same year, it suggests a shared belief that AI demand will require far more infrastructure than exists today. It also suggests a shared risk: if AI demand grows more slowly than expected, or if more efficient models reduce the computing needed per task, all three will own more capacity than they need, at the same time. That is the “AI bubble” argument you will hear in the coming months, and it is not baseless. It is, however, different from saying AI itself is a bubble. A technology can be genuinely transformative and still be overbuilt in the short term, just as railways and fibre broadband were overbuilt before they became essential.

For a small business, the practical difference between AWS, Azure and Google Cloud matters less than the overall direction: cloud capacity is expanding rapidly, AI features are being built into more business software, and the cost of running those features will depend on how efficiently providers can turn their huge capital spending into services that are priced by usage. The providers that succeed will be the ones that can offer the most useful AI capability at a price a normal business can justify, not just the ones that build the most data centres.

What it means for UK small businesses

It is tempting to read a story about $220 billion and a trillion-dollar forecast and conclude it has nothing to do with a firm of twelve people. In fact, it has several practical implications, most of them indirect but all of them worth a few minutes of thought.

First, expect more AI features inside the tools you already pay for, and expect them to be priced in new ways. When a cloud provider builds vast AI capacity, it needs to fill it. One way to fill it is to add AI features to existing products: smarter search, automated summaries, forecasting, customer service assistance, and agent-like tools that can take actions. Some of those features will be included in existing subscriptions, some will be add-ons, and some will be billed by usage, which can be hard to predict. If your team starts using AI features casually, check how they are billed before they become a line item you cannot explain.

Second, review your cloud bill with the same discipline you apply to rent or payroll. AWS pricing is famously detailed, and AI workloads can be expensive in ways that are not obvious at first: data transfer out of a cloud region, storage of large datasets, and always-on AI endpoints that charge even when nobody is using them. A quarterly review that asks three questions, what are we running, what is it costing per month, and could it run on a smaller or cheaper configuration, will catch more waste than any single cost-saving tool. If you do not have someone in-house who can read an AWS bill, that review is a good first task for a consultant, and it is exactly the kind of work we help with.

Third, think about lock-in before you build deeply on one provider. AWS, Azure and Google Cloud all offer excellent services, but moving a complex system from one to another is rarely trivial. If your business is building a product that depends on a specific AI service, ask what would happen if that service changed price, changed terms, or was discontinued. Keeping your data in portable formats, documenting how your system is put together, and avoiding unnecessary dependence on a single provider’s proprietary features are not exciting tasks, but they are the difference between a business that can negotiate and one that cannot.

Fourth, do not confuse Amazon’s spending with a guarantee that AI prices will fall. More capacity can lower unit costs over time, but in the short term, providers need to recover their investment, and demand is strong enough that they have little reason to cut prices aggressively. Budget for AI as a real cost, with a real return to prove, rather than assuming it will become free. A useful test is to ask, for each AI tool you pay for, what task it replaces or improves, and how you would measure that improvement in hours saved, errors avoided, or sales won. If you cannot answer that, you are buying a story, not a tool.

Fifth, consider the energy and location angle, which is increasingly a UK story too. Data centres need power, and the UK has its own debates about where they should be built, how much power they should use, and what that means for local grids. Amazon’s global spending does not answer those questions, but it does mean more data centre capacity is coming, including in Europe, and that UK businesses will have more choice about where their data lives. If your customers care about data residency, or if you work in a regulated sector, ask your provider which region your data is in and what would happen if you needed to move it.

Finally, keep the human side in view. Jassy’s forecast is about revenue, not jobs, and it is easy to read a trillion-dollar cloud business as a story about machines replacing people. The more useful reading is that cloud infrastructure is becoming the factory floor of the AI economy, and the businesses that thrive will be the ones that combine that infrastructure with judgement, customer knowledge, and service that a model cannot copy. A small firm does not need to outspend Amazon. It needs to know its customers better than a generic AI tool does, and to use the tools Amazon and others provide to serve those customers faster and more reliably.

For background on how data centre investment is already reshaping the market, see our coverage of the Morgan Stanley $33 billion data centre funding story, and for a closer look at how companies are trying to measure what their AI spending actually buys, see our post on Anthropic’s AI spending reporting tools.

Honest limits / what to watch

This story has real limits, and a responsible report should name them plainly.

The trillion-dollar figure is a forecast based on growth assumptions, not a result. If AWS grows by 30% instead of more than 40% in 2027 and 2028, the timeline stretches considerably. If a recession reduces enterprise spending, or if a breakthrough makes AI models far cheaper to run, the demand for capacity could grow more slowly than Amazon expects. Forecasts from chief executives on earnings calls are also, by nature, optimistic: they are made by people whose job is to persuade investors that the spending is justified.

The spending itself is not yet proven to earn its return. Amazon says it has a “clear line of sight” to financial returns, but it has not published a detailed breakdown of how much of the $220 billion will be recovered, over what period, or at what margin. Analysts asked about margins on the call for a reason. A business can grow revenue rapidly while making thinner profits, especially if it is discounting to win large AI customers or if power and hardware costs rise.

There is also a concentration risk that deserves attention. When a handful of companies spend hundreds of billions on infrastructure, smaller cloud providers and startups find it harder to compete on capacity alone. That can lead to a market in which a few providers set the terms, which is not necessarily good for price or choice in the long run. Regulators in the UK and Europe are already watching cloud competition closely, and that scrutiny is likely to increase as spending grows.

Energy and environmental questions are not answered by this announcement either. Building more data centres means using more power, and while Amazon and others invest in renewable energy and more efficient cooling, the overall footprint of AI infrastructure is still rising. A small business that cares about its own carbon footprint should ask its provider for real data on energy use and renewable sourcing, not just marketing claims.

And finally, the share price reaction is a reminder that even good news can be read cautiously. A 6% after-hours fall does not mean Amazon is in trouble, but it does mean the market is not willing to take a trillion-dollar forecast on trust. It wants to see the next few quarters of growth, margins and customer demand, and it will judge Amazon on those numbers, not on the ambition alone.

What to watch next

Three things will tell you whether this forecast is on track. The first is AWS growth in the next two quarters: does it stay near 39%, accelerate towards the forecast 40% plus, or slow as customers become more careful with spending? The second is capital spending itself: does Amazon actually spend “meaningfully more” in 2027, and does it explain what that money is buying in terms of regions, chips and capacity? The third is margins: does AWS remain highly profitable while it builds, or does the cost of building eat into profits in a way that forces a change of plan?

Also watch what Microsoft and Google say about their own spending in their next earnings calls. If all three continue to increase spending, the infrastructure boom is a shared industry bet. If one of them slows down, that will be a signal worth reading carefully. And watch for UK-specific announcements: new data centre regions, power deals, or partnerships that bring this global capacity closer to British businesses in practical terms, not just in press releases.

FAQs

What did Amazon say about AWS becoming a $1 trillion business?

On Amazon’s quarterly earnings call on 29 October 2026, chief executive Andy Jassy said AWS could grow by more than 40% in each of 2027 and 2028 and become a trillion-dollar revenue business in the next five to ten years, from a current annualised run rate of $184 billion.

How much is Amazon spending on AI infrastructure?

Amazon expects about $220 billion of cash capital expenditure in 2026, with the majority for AWS, and expects to spend meaningfully more in 2027, mostly to build the data centres, chips and networking that AI workloads need.

How fast is AWS growing now?

AWS revenue rose 39% year on year in the reported quarter, the sixth straight quarter of steady growth on that measure, according to TechCrunch’s report of the call.

Why did Amazon shares fall after the announcement?

Shares fell about 6% in after-hours trading on 29 October, TechCrunch reported, as investors weighed the scale of the spending against the timing of the returns it is supposed to produce.

What does AWS’s spending mean for a UK small business?

In the short term, very little changes overnight. Over time, more capacity should mean more AI services and regions to choose from, but it also means cloud pricing, data transfer charges and vendor lock-in deserve a regular review, not a one-off decision.

Is a $1 trillion AWS a promise or a forecast?

It is a forecast, not a guarantee. Jassy framed it as what Amazon believes is possible if growth stays above 40% in 2027 and 2028, and it depends on AI demand, capacity being built on time, and margins holding up under heavy investment.

CTA: Get a Free AI Consultation

Sources: TechCrunch, AWS will be a $1T business, Jassy says. But first, it will spend big. and Amazon’s quarterly earnings call on 29 October 2026 as reported there.