Amazon Web Services installs a new AI accelerator about every 14 seconds, the company’s chief executive, Andy Jassy, told investors on Amazon’s third-quarter earnings call, and he expects the cloud division to become a trillion-dollar business within five to ten years.
AWS is on track to generate $184 billion in annual revenue this financial year. Third-quarter revenue reached a whisker short of $46 billion, growth of 39.3 per cent year on year, and Jassy said the division should be able to grow by 40 per cent or more in 2027 or 2028.
“It’s important to remember that the vast majority of global IT spend, about 85 percent, is still on premises, and we’re in a 10- to 20-year cycle where that equation will flip to the cloud,” Jassy said. “Conventional wisdom has been that linchpin workloads like mainframes or storage area networks would never move to the cloud,” he added, before noting that “these workloads are moving”.
On the hardware, Jassy said the company installs an accelerator “about every 14 seconds, and almost as fast as we install capacity, customers are gobbling it up”. Its Bedrock AI service is the fastest growing multi-billion dollar service in the company’s history.
Revenue from core compute and storage “grew at their fastest rates in several years”, Jassy said. He argued that the two are linked, because customers running AI workloads want their inferencing near the rest of their data and applications, and that such work “can’t help but drive core AWS demand because AI uses core for work like reinforcement learning, post-training, and tool usage, which can’t help but drive reinvestment back into our broader AWS functionality”. Margins on AI services are already “well ahead” of what core services managed at a comparable stage.
The spending is on a matching scale. Amazon expects $220 billion of capital expenditure in 2026, and spent $75.9 billion last quarter alone, most of it on AWS. “Would it be more convenient if this type of acceleration and opportunity happened over 25 years and you could spread the CapEx over 25 years? Yeah, it might be more convenient, but that’s not the world that we live in, and that’s not the opportunity,” Jassy said.
“We can fund this plan,” he added. “I think it will forever change the size of Amazon. It will forever change our revenue, it will forever change our free cash flow, and we’re going to invest, and we’re going to pursue it to continue to be the significant leader like we are today.”
The rest of the company is growing more slowly. Amazon’s overall quarterly revenue was $180.17 billion, with retail and operations up six per cent on far slimmer margins. Jassy was able to point to Prime Video becoming “a significant business with a revenue run rate in the tens of billions of dollars” and now profitable in its own right. The figures were reported by The Register.
One item in that list is doing more work than the rest. An accelerator every 14 seconds is a rate, and rates can be extended, which is the only thing this newspaper has ever asked of a number.
At one every 14 seconds, the company fits more than four a minute, 257 an hour and roughly 6,170 a day. That is a delivery schedule rather than a procurement strategy. Somewhere there is a loading bay running at the tempo of a bottling plant.
The more revealing comparison is the internal one. AWS grew 39.3 per cent in the quarter. The retail business that gave the company its name grew six per cent. On those trajectories the shop is not a parent company. It is a legacy workload.
The Register put the point in a single line, and it is the right line: if the respective growth rates continue, “in a few years we could be talking about a cloud company with a retail side hustle”. Our analysts would go one step further. The groceries are the on-premises portion of Amazon, and 85 per cent of global IT spend is still, as Jassy observes, on premises.
The capital expenditure confirms the direction of travel. A company that spends $220 billion in a single year against a 10- to 20-year cycle is not buying equipment. It is buying the cycle, and intends to be the thing the equation flips towards.
The Department of Unnecessary Research has one question outstanding. At one accelerator every 14 seconds the installation never stops, and the earnings call lasted considerably longer than 14 seconds. Nobody has yet established how many were fitted while Mr Jassy was describing the rate.