Why I'd Buy Amazon Despite $220 Billion of AI Capex

AWS demand, not management enthusiasm, is driving the spend

Amazon wants to spend roughly $220 billion on AI infrastructure this year, and that is exactly why I find the stock attractive now.

The capex increase points to scarcity, not slack

Amazon lifted 2026 capex to approximately $220 billion. That $20 billion increase was tied to higher memory costs, not a visibly looser strategy. At the same time, AWS revenue reached $42.2 billion and grew 37% year over year, the fastest growth in 18 quarters. Management also said capacity will still fall short of customer demand in 2026 and that the same dynamic is expected in 2027. That is less a capex warning than a sign that supply is trailing demand.

The backlog tells the same story. AWS contracted backlog rose by $132 billion in one quarter to $496 billion, with commitments already extending into 2028. A big buildout is still risky if demand softens, but Amazon does not appear to be building into a void. It is expanding capacity against orders customers have already committed to.

The market's hesitation leaves room for the bull case

The setup also works because doubt is still visible in the tape. Shares are up 4.3% over the last week, but they remain below the May 5 intraday peak. Investors are asking whether Amazon can justify enormous infrastructure spending before the market loses patience. If AWS keeps turning backlog into revenue, Amazon does not need a new story. It just needs the market to give the capex cycle more time.

Backlog and scale make the AI buildout easier to defend

The real question is not whether spending is rising. It is whether customer demand is pulling Amazon forward or management is simply getting more aggressive with capital allocation.

The backlog makes the demand signal clearer

Amazon's answer is straightforward. AWS contracted backlog jumped to $496 billion, with commitments stretching into 2028. That is a stronger signal than a forward-looking narrative. The spending increase also appears tied to higher component costs rather than a dramatically broader project scope. In simple terms, customers are lining up, and the cost of serving them has risen.

That is different from an AI story built mostly on hope. If backlog were soft, the scale of the buildout would be harder to defend. Here, the spending is aligned with signed demand that already spans multiple years.

Amazon's broader business still supports the investment

The other test is whether Amazon can expand AI capacity without stressing the rest of the business. For now, it can. In Q2, Amazon reported total revenue hitting $200.6 billion and operating income rising 43% to $27.5 billion. AWS also remained a large, profitable engine during the buildout.

That matters because Amazon is not funding this capex from a fragile, single-business model. The broader company still has substantial earning power, while AWS continues to generate meaningful profit. That gives Amazon more room to invest in compute capacity and related infrastructure than companies that need the market to believe the spend will eventually work.

The bear case is real, but it is different from no demand

Bulls will argue that this is how a moat widens in AI: secure capacity, lock in customers early, and make supply the constraint. The current evidence supports that view. AWS is growing quickly, backlog is deep, and management said demand remains unsatisfied through 2027 and into 2028.

Bears have a legitimate counterargument too. If fixed costs and depreciation rise faster than profits can absorb them, returns on capital can weaken even as revenue keeps climbing. That is a real risk. But it is different from saying customers disappear. The more visible risk is that Amazon still cannot buy and deploy hardware fast enough to keep up with commitments.

Cash flow is the line that decides whether the story works

The bull case still works, but only if Amazon can turn AI capex into cash before the market stops giving it time. Strong AWS demand and a deep backlog are important, but they are not enough by themselves. At this spending level, investors need to see recoverable cash, not just more infrastructure on the balance sheet.

Free cash flow is the clearest watchpoint

The trap is simple: Amazon can win market share and still disappoint shareholders if free cash flow stays weak for too long. Free cash flow has collapsed to $1.2 billion. That does not prove the buildout is a mistake, but it does show that the payback period may not be painless.

The pressure is easier to see in peer context. Amazon is part of a mega-cap AI spending race where about $610 billion in capital expenditures is being planned across four companies. In that kind of race, the question is not only who is spending the most, but who gets cash back first.

What has to hold in the next few quarters

Management also left the market a simple scoreboard. Amazon guided Q3 revenue of $197 billion to $202 billion and operating income of $22.5 billion to $26.5 billion. If those ranges hold while the AI buildout continues, Amazon likely has enough time to monetize the spend. If earnings slip while capex stays elevated, the narrative can shift quickly from moat-building to excessive capital intensity.

My stance is to buy dips only if cash flow shows signs of improving. The main invalidation is simple: free cash flow stays weak while spending remains near the current pace. That is the line between a durable moat being funded and a growth story running out of patience.