AWS Is Sold Out Through 2027. Here's the $220B Catch.

Amazon's CEO says its cloud has no spare capacity for 2026, very little for 2027 and some already spoken for in 2028. Demand is not the question anymore. The question is what it costs to meet it, and Thursday's earnings are where investors get the first answer.

HeyTheo Research · Monday, October 26, 2026

Quick Read

• Andy Jassy said AWS capacity for 2026 is sold out, 2027 is "largely sold out" and "a good chunk" of 2028 is already spoken for.

• AWS grew 37% last quarter, its fastest pace in years, and its order backlog roughly doubled in six months to about $496 billion.

• The catch is the bill. Amazon plans about $220 billion of capital spending in 2026, and its free cash flow over the past year turned negative for the first time in years.

• Amazon (AMZN) reports on Thursday, Oct 29, after the close. Four numbers decide how the market reads "sold out": AWS growth, backlog, the spending plan and cash flow.

What Jassy said

In a podcast interview published last week, Jassy put dates on a shortage the industry has talked about for a year.

"2026 is obviously, at this point, sold out. '27 is largely sold out." Then: "A good chunk of '28 is sold out." AWS keeps some capacity back for on-demand customers and its Bedrock AI service, so "largely" does not mean every server is taken.

He also named the bottleneck: "There's only so much data center and power today. There's only so much memory." And he does not expect that shortage to end in 2027.

Sold out sounds like good news, and in one sense it is. A business that cannot meet demand has pricing power and years of visibility. But a cloud provider can only sell capacity it has already built, and building it is the most expensive thing Amazon has ever done.

The demand side: how fast AWS is growing

Source: Amazon filings and earnings calls, Drillr

Growth sped up for four straight quarters. For a business running at about $169 billion a year, that is unusual: most cloud businesses slow down as they get bigger.

The backlog matters more than the growth rate. It is the value of contracts customers have signed but AWS has not yet delivered. It was about $244 billion at the end of 2025, about $364 billion by the first-quarter call, and about $496 billion at the end of June.

That is roughly three years of today's AWS revenue already under contract. "Sold out" is not a figure of speech.

Research cue · try it in HeyTheo

Ask Theo: "What does Amazon's $496 billion backlog mean for AWS revenue over the next three years?" Then ask for the bull and bear case on Amazon if AWS growth holds near 37%. You get a first pass with the assumptions laid out, so you can check them yourself. Open HeyTheo

Start with: search AMZN in the app.

The supply side: what "sold out" costs

Here is the other half of the story.

Source: Amazon filings, Drillr

Amazon now expects about $220 billion of capital spending in 2026, up from about $200 billion. Jassy blamed the increase on one thing: the higher cost of memory chips. That is roughly $600 million a day.

Three things follow from that number:

1. Cash flow goes negative before it comes back. Amazon earns more cash than ever, but it is spending more than it earns. Jassy has said to expect free cash flow headwinds until the new data centers start earning.

2. Depreciation is coming. Servers wear out in about five or six years, and that cost hits profits every year. AWS depreciation rose 67% in the second quarter, to about 19% of AWS revenue. The margin has held so far because prices and usage are rising faster.

3. Suppliers win either way. Every dollar Amazon spends goes to someone. Memory makers like Micron (MU), chip companies like Nvidia (NVDA), and the companies supplying power and data center equipment collect it whether or not the payoff arrives on time. We covered the chip side in our Amazon TorchNeuron note and the memory squeeze in our Nvidia, Samsung and TSMC memory note.

Research cue · try it in HeyTheo

Build a basket of the AI spenders and their suppliers: Amazon, Microsoft, Alphabet, Nvidia and Micron. Then set a trigger to flag any name that moves more than 5% after earnings this week. One alert instead of five tabs. Open HeyTheo

Start with: create a basket named "AI capex week."

Rather read first? Our free daily brief lands in your inbox before the open. Send it to me

The bull case and the bear case

The bull case. Capacity sold out three years ahead means revenue already under contract. Higher prices for scarce compute support margins, the backlog covers years of growth, and Amazon's own Trainium chips can lower its cost per server over time. If AWS keeps growing in the high 30s, the spending pays for itself, and free cash flow recovers sharply once the build slows.

The bear case. Backlog is a promise, not cash, and much of it may come from a handful of large AI customers whose own funding depends on markets staying open. We looked at that credit risk in our CoreWeave and Oracle debt note. Memory and power costs keep rising, so each new server costs more than the last. Debt has roughly doubled this year. If AI demand cools before the bills are paid, Amazon is left with expensive capacity and years of depreciation.

Both cases agree on one thing: the next 12 months are the most expensive part of the cycle.

What to watch on Thursday

Source: Amazon guidance and filings, Drillr

Amazon guided to $197 billion to $202 billion in sales for the quarter. Analysts are near the top of that range, with earnings estimates around $1.96 to $2.01 a share.

Context matters too. Microsoft (MSFT) and Alphabet (GOOGL) report on Wednesday, Oct 28, hours after the Fed's rate decision. Their cloud numbers will set expectations before Amazon speaks. If all three say capacity is the limit, the debate moves from "is AI demand real" to "who can build fastest."

Research it yourself: rules HeyTheo tracks

• AI capex basket. Amazon, Microsoft, Alphabet and their main suppliers tracked side by side through earnings week.

• Earnings move trigger. Flags when a stock moves more than 5% the day after results.

• Backtester. See how Amazon traded after its past earnings reports when AWS growth sped up versus slowed.

• Ask Theo. Try: "Compare AWS, Azure and Google Cloud growth over the last four quarters" or "What is the bull and bear case for Amazon if capital spending goes above $220 billion?"

Check the rule behind any trigger before acting on it. You trade through your own broker; HeyTheo helps you decide.

The Waitlist Economy

Most companies worry about finding customers. Amazon's cloud has the opposite problem: a waiting list that runs into 2028.

That flips the usual question. Investors no longer need to ask whether AI demand is real. They need to ask whether a company can afford to meet it, and how long it takes for $220 billion of servers to start paying for themselves. Thursday is the first checkpoint.

Do your own first pass before Thursday. Create a free HeyTheo account by Oct 31 and get a month of Pro free. No card needed. On the pricing page, tap Sign up for Free till Oct 31. See the offer

Prefer the phone? Get the app

Reading on a laptop, or not ready to install? Get the free daily brief by email. Get the daily brief

Sources

Amazon, a16z, Drillr, Yahoo Finance, Seeking Alpha, StockTitan, Shattered, TechFlow

Disclaimer

Disclaimer: HeyTheo is a research and education platform, not an investment adviser or broker-dealer. Nothing here is advice to buy, sell, or hold any security. You trade through your own broker; HeyTheo helps you decide. Backtested results are hypothetical and do not guarantee future returns. References to governments, officials, or policies are for market context only and are not political endorsements. All investing involves risk, including loss of principal. Data is as of the dates noted.