Copyright Investing.com
When (NASDAQ: ) announced it would spend roughly $125 billion on capital expenditures in 2025 (and even more in 2026), Wall Street responded with an 11% stock surge. When (NASDAQ: ) announced similar massive AI infrastructure spending just hours earlier, investors punished the stock with a 12% collapse. Same industry, same AI narrative, opposite outcomes. For small investors trying to make sense of Big Tech’s AI arms race, this divergence reveals something critical that’s hiding in plain sight: Amazon may be the only company monetizing its AI infrastructure twice, once by selling it to customers through AWS, and again by using it internally to boost its retail and advertising businesses. This structural advantage explains why Amazon’s capex binge looks like smart investing while Meta’s looks like a gamble. The Surface Story: AWS Revenue Makes Capex Palatable The most obvious explanation for Amazon’s stock surge is simple: AWS grew 20% to roughly $33 billion in Q3 2025, its strongest growth rate since 2022. Unlike Meta, which generates 98% of revenue from advertising, Amazon is building AI infrastructure it can immediately sell to enterprise customers. AWS already carries 30% operating margins, meaning every dollar of infrastructure spending generates profitable revenue within quarters, not years. Meta, by contrast, is asking investors to fund $72 billion in AI capex while acknowledging that current AI initiatives generate zero revenue. The company’s CFO even admitted "advertising is no longer the priority" while simultaneously warning that capex would be "substantially larger" in 2026. For investors, that’s a bet on unproven future returns, the exact narrative that burned them on the $70 billion metaverse experiment. Here’s where most analysis stops. However, the real story goes much deeper. The Hidden Play: Internal Subsidization Through Transfer Economics Amazon is building much more than data centers to sell AWS capacity to Microsoft, Anthropic, and thousands of enterprise customers. It’s also making infrastructure that its own retail, advertising, and logistics operations can use at cost, not at the market rates it charges external customers. Let’s think about what this means in practice: When Amazon builds a new AI chip cluster for $1 billion, it can: Sell compute capacity to external AWS customers at full market rates (generating external revenue), and Use the same infrastructure internally to power product recommendations, dynamic pricing algorithms, fraud detection, and advertising optimization (reducing internal costs and improving margins) The second benefit is almost invisible in financial statements but potentially huge. Amazon’s retail business operates on razor-thin margins, often 3-5% operating margins for online stores. If AI-powered systems can improve inventory forecasting by 2%, reduce return rates by 1%, or increase advertising click-through rates by 5%, those efficiency gains flow directly to the bottom line across a $280 billion retail operation. This is double monetization: the same capex dollar generates both external revenue (AWS sales) and internal cost savings (retail efficiency). Revenue Diversification: Meta generates 98% of revenue from advertising. Amazon generates just 9% from ads, with revenue spread across online stores (39%), third-party seller services (24%), AWS cloud (17%), advertising (9%), subscriptions (7%), and other services (4%). Why Meta, Google, and Microsoft Can’t Replicate This Let’s compare Amazon’s position to its Big Tech peers: Meta doesn’t sell cloud infrastructure at scale, so every AI dollar is a pure cost that must be justified by future advertising improvement or new product revenue. There’s no external customer paying Meta to subsidize its AI experimentation. To be fair, some Meta bulls argue the company is playing the long game, spending aggressively now to dominate AI infrastructure the way it dominated social networking, even if that means years of losses before monetization. The bet is that whoever builds the biggest, fastest AI infrastructure will own the next platform, and customer adoption will follow. But there’s one critical difference: Meta’s social networking dominance came with a clear monetization path (advertising) that proved out within 3-4 years of launch. AI infrastructure spending is now in year three with no comparable revenue stream visible. Amazon, by contrast, doesn’t need to wait for a future payoff. It’s monetizing infrastructure today through AWS while at the same time positioning for tomorrow’s AI applications. Google (NASDAQ: ) and Microsoft (NASDAQ: ) do have cloud businesses (Google Cloud and Azure), so they can monetize infrastructure externally. But neither has Amazon’s integrated retail-plus-cloud model. Google doesn’t run a $280 billion e-commerce operation that benefits from the same AI infrastructure it sells. Microsoft doesn’t operate a global logistics network that can be optimized with AI chips it’s already purchased for Azure. For this reason, Amazon uniquely combines: A massive external cloud business (AWS) that generates revenue A massive internal retail/logistics operation that consumes compute Advertising and subscription services that benefit from the same AI infrastructure This creates structural pricing power that competitors lack. Amazon can afford to price AWS competitively because it’s recovering value internally even if external margins compress. Meta has no such cushion. The Questions Small Investors Should Be Asking If you’re evaluating Big Tech’s AI spending, here are the critical questions about this hidden dynamic: 1. Is the company selling the infrastructure or just using it? Amazon sells it (AWS). Meta only uses it internally. That difference is worth billions in de-risked capex. 2. Can we quantify the internal efficiency gains? This is the hard part, and where small investors have leverage. Amazon doesn’t break out how much AWS capacity is consumed internally versus sold externally. The company doesn’t report transfer pricing between AWS and its retail segments. These are questions for earnings calls and shareholder meetings. Specifically, investors should ask: What percentage of AWS capacity is used internally versus sold to external customers? How is internal AWS usage priced in segment reporting? What cost savings has AI-powered infrastructure delivered to retail operations in the past year? 3. Is advertising growth sustainable or inflated by AI subsidies? Amazon’s advertising revenue grew 24% to $17.7 billion in Q3. That’s impressive, but is some of that growth due to better ad targeting powered by AI infrastructure that’s being depreciated across AWS, not advertising? If so, the advertising segment’s profitability might be understated, and the true ROI of AI capex is higher than it appears. 4. What happens if AWS growth slows? If AWS reaccelerates to 20%+ growth, Amazon’s double-monetization thesis strengthens. But if AWS growth decelerates back to 13-15% (as it did in 2023-2024), does Amazon’s capex become a liability? The answer depends on whether internal efficiency gains are large enough to justify spending even without external AWS revenue growth. The Capex Coverage Test: A Simple Framework Here’s a simple way to evaluate whether AI capex is sustainable across Big Tech: External Revenue ÷ Estimated Annual Depreciation For Amazon’s AWS: Q3 revenue: $33 billion (annualized ~$132 billion) Estimated annual depreciation on $125B capex: ~$30 billion (assuming 3-5 year chip life) Coverage ratio: 4.4x For Meta’s AI infrastructure: Current AI-specific revenue: $0 billion Estimated annual depreciation on $72B capex: ~$22 billion Coverage ratio: 0x Amazon’s infrastructure pays for itself through external sales alone, before counting any internal benefits. Meta’s doesn’t. That’s the divergence in one number. Why This Matters Now: The 2026 Capex Reckoning Both Amazon and Meta have signaled that 2026 capex will be even higher than 2025. For Amazon, that’s a sign of strength. AWS demand is accelerating, and the company can fund growth from operating cash flow (which rose 16% to $130.7 billion TTM). For Meta, it’s a red flag. Advertising growth is uncertain, and rising capex without proven AI monetization could turn free cash flow negative. Small investors have a narrow window to position themselves before this dynamic becomes consensus. The market is still treating Big Tech AI spending as a monolithic story ("everyone’s spending big on AI"). But the underlying economics are very different. Amazon is building infrastructure it sells and uses. Meta is building infrastructure it only uses. That’s the difference between a self-funding growth engine and a bet that requires everything to go right. What This Means for Small Investors If you’re evaluating tech stocks for AI exposure, ask yourself: Who’s monetizing the infrastructure, not just the applications built on top of it? Amazon is monetizing twice, externally through AWS and internally through retail efficiency. That double-layered return makes its $125 billion capex bill one of the safest AI bets in Big Tech. Meta, despite its strengths in AI research and model development, is still searching for its first monetization path. Until it finds one, every capex dollar is a leap of faith. For small investors, the lesson is clear: Follow the revenue, not the rhetoric. In the AI infrastructure race, Amazon isn’t just spending more. It’s spending smarter. And that structural advantage is worth paying attention to.
