Wednesday, November 11, 2026
Privacy-First Edition
Back to NNN
Business

AWS CEO: Amazon Lands “One GPU Every 14 Seconds” and Still Expects to Be Short of Capacity

Via TIKR

AWS CEO: Amazon Lands “One GPU Every 14 Seconds” and Still Expects to Be Short of Capacity Chris Arnold Wed, November 11, 2026 at 10:43 AM EST 6 min read AMZN -1.44% Trade AMZN on Coinbase Trading disclosure Trading disclosure The above button links to Coinbase. Yahoo Finance is not a broker-dealer or investment adviser and does not offer securities or cryptocurrencies for sale or facilitate trading. Coinbase pays us for certain activity generated through this link. Prices displayed are informational.

AWS CEO Matt Garman says Amazon is installing a GPU every 14 seconds and still expects AWS to be short of capacity for the next couple of years.

With demand ahead of supply and local pushback slowing new data centers, the capacity AWS already runs is worth more.

AWS brought in about 23% of Amazon's third-quarter revenue but about 63% of its operating income, and analysts expect normalized EPS to nearly double from 2025 to 2028.

The cost shows up in cash: free cash flow was a $42.4 billion outflow over the 12 months to Sept. 30, and Amazon's 2027 capex outlook in early February is the next test.

Amazon (AMZN:NASDAQ) is adding AI hardware at a pace that's hard to picture, and the head of its cloud business says it still isn't enough.

On this week's episode of Sources with Alex Heath, Heath asked AWS CEO Matt Garman about building data centers in space. Garman called it "highly unlikely to be economically feasible or practical in the next decade or two," partly because the whole world doesn't manage one rocket launch a day.

Then he offered a comparison: "we're landing in a data center today, one GPU every 14 seconds."

That still doesn't meet demand. "Right now the demand is far outstripping the supply," Garman said earlier in the conversation, and even with all the capex Amazon is spending, "I anticipate that we'll be short of capacity for the next couple of years."

So the most profitable part of Amazon is sold out.

TIKR tracks 61 analysts on Amazon. How many say "Buy"?

See the full breakdown of Buy, Hold, and Sell ratings on Amazon, along with Wall Street's revenue and earnings estimates, all in one place. More than 500,000 investors use TIKR to research stocks like this one. See Amazon's Analyst Ratings

Garman went further. "I could choose to sell out every single accelerator that I have that I know is landing in 2027 and just be done with it," he said. AWS doesn't, because it holds some capacity back for enterprises and startups that come along later.

The demand also comes with contracts behind it. Garman said most of AWS's larger customers commit to their spending for three to six years, some for as long as 10, and they often end up spending more than they committed. Amazon put AWS's backlog at $578 billion at the end of September.

Garman isn't worried that cheaper AI will shrink the business, either: "every time we see inference get, you know, 50% cheaper, we see an order of magnitude more usage."

The midterms make this matter more. On Nov. 3, Democrats won the House and the Senate, and ten Ohio towns banned large data centers. Two of those bans reportedly affect Amazon. Across the country, the results point more to slower permitting than to outright bans, which could force hyperscalers into capex discipline.

If every new data center takes longer to approve, capacity that's already running and already sold is worth more. AWS has more of it than almost anyone, though it needs new sites too.

In the third quarter, AWS sales rose 39% to $46.0 billion. Its $17.8 billion of operating income was about 63% of Amazon's $28.1 billion total, on about 23% of Amazon's sales, and its operating margin rose to 38.7% from 34.6% a year earlier. On Amazon's third-quarter call, CEO Andy Jassy talked about AWS growing 40% a year in each of the next two years, and said it could "ultimately end up as a trillion-dollar-plus business."

Analysts expect Amazon's normalized EPS to rise from $7.17 in 2025 to $13.99 in 2028.

AWS's growing share of profits is a big reason Amazon's operating margin has roughly doubled since 2021, from 5.3% to 11.2% in 2025.

Of course, a cloud that can't keep up with demand keeps spending to catch up. In 2025, Amazon's $132 billion of capex used up nearly all of the $140 billion its operations brought in.

The spending has only grown since. Over the 12 months to Sept. 30, net capex reached $210.7 billion, and free cash flow went from a $14.8 billion inflow a year earlier to a $42.4 billion outflow. Management expects about $220 billion of cash capex in 2026, and long-term debt doubled to $128.9 billion in the first half of the year. Analysts don't expect free cash flow to turn positive again until 2028.

The shortage is also why the spending should pay off. Management says servers break even in a little under three years and last at least five to six years, and most AI capacity is contracted for at least five years. A server that's sold before it's plugged in and pays for itself in under three years is a good place to put cash.

The risk is who's doing much of the buying. OpenAI and Anthropic each added $100 billion or more of AWS commitments this year, and Amazon has invested $68 billion in the two, so part of that backlog depends on labs Amazon helps fund.

The way I see it, Garman just gave investors the best reason to put up with Amazon's spending. A business that could sell out its 2027 chips today and earns about 63% of Amazon's operating income has room to grow into a $220 billion capex budget, especially if local pushback slows everyone else's build-out. And investors are paying less for those earnings than they were: Amazon's P/E has fallen from 34 to 21.

The next test comes in early February, when Amazon gives its 2027 capex outlook with fourth-quarter results. If capex keeps growing faster than AWS's revenue, investors will be waiting past 2028 for that free cash flow.

TIKR lets you forecast the future price of any stock in less than a minute. Just enter a few assumptions into TIKR's valuation model and see what Amazon could be worth. Start from Wall Street consensus estimates, or adjust the inputs to reflect your own view of the business. It's free to use.

Read original at Yahoo Finance News

The Perspectives

0 verified voices · Three viewpoints · Real discourse

Left
0
Be the first to share a left perspective
Center
0
Be the first to share a center perspective
Right
0
Be the first to share a right perspective

Related Stories