Wall Street Lets Amazon AI Spending Burn $220 Billion

Amazon AI spending is getting the treatment most companies want and few receive: investors are looking past a cash drain because AWS is already turning the AI buildout into revenue.

Amazon reported better-than-expected second-quarter earnings Thursday, with net sales up 20% and AWS revenue up 37% year over year to $42 billion, according to TechCrunch. The stock jumped nearly 10% in after-hours trading, even as Amazon raised its 2026 capex forecast from $200 billion to $220 billion.

That reaction says more than the headline numbers. Investors are not rewarding AI spending in general. They are rewarding AI spending when it sits inside a cloud platform that can rent out capacity, bundle services, and show demand before the full infrastructure bill comes due.

Wall Street is giving Amazon AI spending a pass because AWS can turn capacity into rent

The tension is blunt. Amazon is spending heavily on data centers at the same time its cash position is tightening, but investors shrugged because AWS gave them a revenue story.

Amazon spent $173 billion on property and equipment for the fiscal year ended June 30, up from $107.65 billion a year earlier. TechCrunch notes that this category includes GPUs, natural gas turbines, and plots of land. That is not a light expansion cycle. It is a balance-sheet commitment to the physical layer of AI.

Under normal conditions, that kind of spending would invite harsher treatment. Amazon also ended the quarter with $7.6 billion less cash than it had 12 months earlier, and TechCrunch says this marked its first period of negative free cash flow this year.

The reason investors tolerated it is simple: AWS is not an abstract AI bet. It is already producing large cloud revenue while demand for AI compute is rising alongside supply. XOOMAR analysis: the market is treating Amazon less like an AI product company and more like an AI landlord. The rent is not theoretical.

For more context on Amazon’s broader AI ambitions, see XOOMAR’s earlier coverage of how $200 Billion Sales Let Amazon AI Agents Invade Workflows. That matters here because infrastructure spending only works if customers keep finding reasons to consume more compute.

---

The numbers behind Amazon’s data center binge and the investor shrug

Amazon’s spending profile is aggressive even by hyperscaler standards. The company lifted its 2026 capex forecast by $20 billion, from $200 billion to $220 billion, while continuing to pour money into assets tied to cloud capacity.

The investor reaction shows which number mattered most:

The key economic question is utilization. Amazon is building capacity before all of it can be sold, and TechCrunch points to the years-long lag between breaking ground on a data center and selling that capacity. That lag is the risk. AWS growth is the reassurance.

XOOMAR analysis: investors appear willing to finance the gap because AWS has a proven channel to market. If AI training and inference demand keeps rising inside AWS, the spending can look like inventory for a high-demand business. If demand softens, the same spending becomes stranded capacity with a very expensive power bill.

Amazon gets a different AI multiple than companies still proving demand

Amazon’s advantage is not that it spends less. It is that it can point to a business line already monetizing the AI infrastructure wave.

The market is separating AI infrastructure with visible revenue from AI spending that still lacks a clear payback path. Amazon’s after-hours move suggests investors are giving more credit to companies that can connect heavy infrastructure outlays to current cloud demand.

“We see the AI business following very much the same margin trajectory we saw in the core business before,” Jassy said during the company’s Q2 earnings call.

That quote is the strategic center of Amazon’s pitch. The company is arguing that AI can follow a familiar cloud pattern: invest heavily upfront, then let customers consume capacity and services as demand matures.

XOOMAR analysis: this is why cloud hosts are getting a cleaner AI narrative than many AI startups. The lab or app company has to prove pricing, retention, and margins. The cloud host gets paid when those companies train, test, deploy, and scale. That does not remove demand risk. It moves it one layer down the stack.

Investors have shown similar selectivity across earnings. XOOMAR has been tracking that split in Two Earnings Shocks Split Dow Jones Near Record High, where earnings reactions turned heavily on whether companies could connect spending to durable cash generation.

---

Amazon wants the cloud playbook, not an infrastructure hangover

The best version of Amazon’s AI capex story resembles the cloud buildout: hyperscalers spend ahead of demand, customers move workloads onto rented infrastructure, and utilization catches up over time.

The dangerous version is different: there may not be enough durable demand to justify the scale of the AI infrastructure buildout. If customers do not turn AI usage into revenue, the cloud bills supporting that buildout become harder to sustain.

That is the right question because Amazon’s revenue is someone else’s cost. In cases where AI labs buy massive cloud capacity, one company’s growth can depend on another company’s willingness and ability to keep spending.

This is the part investors may be underpricing. Cloud hosts look safer than model labs because they sell the picks and shovels. But if too many customers cannot turn AI usage into their own revenue, the bills eventually hit resistance. Amazon is a few steps removed from that problem, not immune to it.

AWS customers and Amazon shareholders are buying different sides of the same bet

For shareholders, the bullish case is clear. Amazon is spending to keep AWS central to enterprise AI, and the latest quarter gave them evidence that cloud revenue is growing fast enough to justify patience.

The skeptical case is just as clear. Capex is rising, cash is falling, and the AI buildout has no obvious finish line. A company can be strategically right and still overbuild at the wrong price.

For AWS customers, the appeal is access. They do not need to own the GPUs, land, turbines, or data center projects sitting inside Amazon’s property and equipment line. They can rent capacity through AWS and use services like Bedrock. But XOOMAR analysis: their willingness to keep doing that depends on whether AI workloads move from experimentation into sustained production use.

For Amazon’s rivals, the same investor test likely applies. Strong cloud revenue earns patience. Heavy spending without a clearly attached revenue engine gets punished.

Amazon’s AI spending honeymoon depends on durable AWS revenue

Amazon has earned more trust than most AI spenders because AWS gives investors a visible path from capital spending to revenue. That is why the market accepted a nearly 10% after-hours stock jump alongside higher capex and weaker cash flow.

The honeymoon will last only if the evidence keeps lining up:

• AWS growth: Revenue needs to keep showing that demand is rising with capacity.
• Cash pressure: Investors may tolerate negative free cash flow periods, but only if the payoff becomes clearer.
• AI usage: Training and inference demand must become durable, not just bursty.
• Customer economics: AI labs and enterprise users must afford the bills that become Amazon’s revenue.
• Efficiency: Scrutiny should shift from how much Amazon spends to how well that spending converts into AWS returns.

The practical read is narrow but powerful. Investors do not love AI spending by default. They love AI spending when it is attached to a cloud tollbooth with paying traffic.

Amazon has that tollbooth. Now it has to prove the traffic is permanent.

The Bottom Line

• Investors are rewarding AI spending when it is tied to immediate cloud revenue, not speculative promises.
• AWS gives Amazon a clearer path to monetize AI infrastructure through compute rentals and bundled services.
• Amazon’s rising capex and weaker cash position show how expensive the AI infrastructure race has become.

---

Originally published on XOOMAR. For more news and analysis, visit XOOMAR.