Big tech earnings preview: Can Google, Meta, Amazon, and Microsoft turn AI CapEx into durable revenue?

Big Tech Earnings: Can AI CapEx Convert to Revenue?

Four of the seven largest companies report earnings today—Google, Amazon, Microsoft, and Meta—representing just under 20% of the S&P 500's total market cap. The earnings collectively hinge on a single question: can hyperscalers turn massive AI infrastructure spending into durable revenue before depreciation catches up?

The context is stark. Amazon guided to $200 billion in CapEx for 2026. Google, Microsoft, and Meta are all spending at least $100 billion annually. Semi Analysis expects hyperscaler CapEx guidance to be revised upward, with the market betting that cloud revenue acceleration and positive ROI on infrastructure investments justify the spending. But investors want proof of the conversion—not just cash burn, not just capex stories, but a documented path from hardware spending to sustainable profit.

The core tension: these companies can keep drawing down cash endlessly, but at some point they need to stop making money entirely if the revenue doesn't materialize. The question is how quickly the cash converts back.

Microsoft: The Clearest Enterprise Read

Microsoft reports with the strongest signal on enterprise AI monetization. Azure is growing at 39%, the largest single lever on CapEx running at roughly $150 billion annually. The most eye-popping number from last quarter was remaining performance obligations at $625 billion, up 110%—a clear signal that companies have signed long-term compute contracts ahead of demand. About 45% of that RPO comes from OpenAI partnerships, but dozens of other vendors have committed to multiyear deals at scale.

The earnings will surface three critical data points. First, gross margins on Microsoft Cloud. If customers push back on pricing and demand discounts as AI commoditizes, margin compression would show up here. Second, Copilot adoption and ARPU—how many customers are paying incremental dollars for AI-powered features, and how much are they paying? GitHub Copilot hit $500 million ARR very quickly, but the conversation in developer tools has been dominated by startups like Cursor, Windsurf, and others. Today's numbers will reveal whether Microsoft's enterprise distribution machine can press GitHub Copilot forward even if it's not the most hyped product. Third, M365 seat growth. Are companies adding seats for human workers, or are AI agents consolidating headcount? The answer matters for the entire SaaS model.

Microsoft also has the cleanest view into how AI flows through the American economy. It knows what APIs are generating token spend, what enterprise workflows are adopting Copilot, and how much customers are willing to pay for integrated AI. That visibility is valuable for the market.

Google: The Integrated Stack Question

Google has the most fully integrated AI stack—consumer distribution through Search and YouTube, model training at DeepMind, custom chips with TPU, and product surfaces to deploy features across Workspace, Search, YouTube, Android, and Cloud. The flywheel should spin fast. But investors are asking whether AI overviews and Gemini monetize fast enough to offset potential declines in search ad revenue. Search is still working well, and Cloud offers tons of growth headroom, but the key question is unit economics: are AI search features expanding usage and ad ROI, or compressing the financial model?

Amazon: AWS Acceleration and Cash Flow

Amazon guided $200 billion CapEx for 2026 and reported AWS growing at 24% in Q4, with $21.3 billion in ads revenue—an almost $100 billion annual run rate that provides substantial cash flow to fund infrastructure buildout. Free cash flow hit $11.2 billion last quarter, down due to AI spending increases.

Consensus expects AWS revenue around $36.7 billion with mid-20s growth, but the market is hoping for 30% acceleration. If AWS accelerates meaningfully, all the CapEx looks prescient—buying scarce compute capacity ahead of demand at a time when being compute-rich is strategically valuable.

Meta: AI Works Immediately

Meta faces less pressure than its peers. The company grew revenue 24% to nearly $60 billion last quarter, with 3.8 billion daily active users. Ad impressions rose 18%, average price per ad rose 6%. Family of apps operating income hit $30.8 billion, making the $6 billion loss at Reality Labs seem almost immaterial—a side bet on future platforms.

Meta's advantage is directness. When a new AI model improves ad placement effectiveness, Meta can AB test it across the entire family of apps within days. No diffusion lag, no enterprise sales cycle, no change management. A small improvement in ad targeting can move the needle by $20 billion in profit. The company has no real diffusion problem. If there's a breakthrough in reasoning or ranking, it shows up in ad performance immediately. Consumers don't notice the change; they just see more relevant ads.

CapEx guidance jumped from $72.2 billion last year to $115–$135 billion this year—near a doubling. Expectations imply 31% revenue growth. The math works if AI-driven ad improvements drive incremental growth faster than depreciation erodes returns.

---

The overarching thesis unifies all four: hyperscalers are betting that AI creates proprietary distribution, measurable customer retention, and revenue growth that materializes before infrastructure costs depreciate. Each company has a different pathway—Microsoft through enterprise contracts, Google through integrated products, Amazon through compute scarcity, Meta through immediate ad performance. Today's earnings will test whether any of those pathways hold.