Key Takeaways
- Combined AI capital expenditure guidance for Microsoft, Alphabet, Meta, and Amazon now exceeds $725 billion for 2026
- Google Cloud revenue grew 63% year-over-year to $20 billion, driven by generative AI demand
- Microsoft's AI revenue has reached a $37 billion annualized run-rate, up 123% from last year
The 'Prove-It' Quarter Delivered
For two years, Wall Street tolerated massive AI infrastructure spending on faith. That faith got tested in Q1 2026. Apple, Amazon, Meta, Google, and Microsoft all reported earnings last week. The numbers answered the central question: is the spending working?
It is. First-quarter capital expenditure from the four hyperscalers (excluding Apple, which runs a different model) exceeded three times the inflation-adjusted cost of the Manhattan Project. And unlike speculative bets of previous cycles, the revenue showed up alongside the spending.
“The 'AI capex is speculative' narrative is dead... This was the prove-it quarter. They proved it.”
— Daniel Newman, CEO of The Futurum Group
Google and Meta: Same Great Numbers, Different Reactions
Ben Thompson at Stratechery flagged an odd pattern in market reactions. Google posted excellent numbers. Meta posted arguably better numbers. Wall Street loved Google. Meta stumbled.
Google Cloud hit $20 billion in quarterly revenue, a 63% year-over-year jump. That growth rate is a record for the division, driven almost entirely by enterprises adopting generative AI tools. Sundar Pichai didn't undersell it.
“2026 is off to a terrific start. Our AI investments and full stack approach are lighting up every part of the business... we are compute-constrained in the near term.”
— Sundar Pichai, CEO at Alphabet
Meta, meanwhile, has built a $10 billion annual revenue run-rate from AI-generated video ads through its Advantage+ system. The product works. Advertisers get better returns. Meta gets more ad dollars. The market still punished the stock on spending concerns.
Thompson's read: investors are still skittish about Meta's infrastructure-first strategy even when it produces results. Google benefits from cloud revenue being an obvious line item. Meta's AI gains are baked into advertising, harder to isolate, easier to doubt.
Microsoft: Demand Outstripping Supply
Microsoft's AI story is the simplest to track. The company now reports AI revenue separately. It hit a $37 billion annualized run-rate, up 123% from the same period last year. Copilot subscriptions, Azure AI services, and enterprise deals are all contributing.
CFO Amy Hood delivered the most telling quote of earnings season: the company is capacity-constrained through at least the end of fiscal 2026. Demand continues to outstrip available supply. That's not a spending problem. That's a scarcity problem.
For context, Microsoft spent aggressively on data centers through 2024 and 2025. The company has more GPU capacity than almost anyone outside of hyperscalers. It still can't keep up.
Amazon's Infrastructure Strategy
Thompson's Tuesday article connected Amazon's AI strategy to its longer infrastructure history. The company has always spent early and big on physical capacity. Warehouses. Delivery networks. AWS data centers. AI follows the same playbook.
Amazon is betting on custom silicon (Trainium chips for training, Inferentia for inference) to reduce long-term costs. The strategy takes longer to pay off than buying Nvidia GPUs. It also creates a structural cost advantage if it works.
AWS reported solid AI growth, though the company doesn't break out AI revenue as cleanly as Microsoft. What's clear: Amazon is playing the same game it's always played. Invest in infrastructure before demand proves out. Reap margins later.
Apple: The Exception
Apple doesn't fit the pattern. The company doesn't operate hyperscale cloud infrastructure. It doesn't sell AI services to enterprises. Its AI strategy runs through device features and partnerships (notably with OpenAI for Apple Intelligence).
Thompson covered Apple's Q1 separately, noting the company's different business model. Apple's capital expenditure goes to retail, manufacturing, and R&D. Not data centers. The company is an AI consumer, not an AI infrastructure provider.
What the Numbers Actually Mean
Thompson's core argument: the spending looks irrational only if you ignore the revenue. Google Cloud at 63% growth justifies infrastructure investment. Microsoft at $37 billion AI run-rate justifies data center expansion. Meta at $10 billion in AI-generated ad revenue justifies its model training costs.
The shift from the 'LLM hype phase' to what Thompson calls the 'agentic era' is real. Companies aren't building chatbots anymore. They're integrating autonomous AI agents into enterprise workflows. That requires more compute, not less.
The skeptic's case (AI spending will hit a wall, returns will diminish, the bubble will pop) isn't impossible. But Q1 2026 didn't support it. Revenue grew. Demand exceeded supply. The companies spending the most are also the companies earning the most from AI.
Beyond the Numbers
This week's Stratechery also featured an interview with Wall Street Journal tech columnist Joanna Stern about her new book on AI. Stern discussed using an LLM to make a career change, how AI is transforming medical diagnostics (including mammograms), and the limits of current models.
One highlight: ChatGPT apparently misdiagnosed a praying mantis pregnancy. A good reminder that even $725 billion in spending hasn't solved everything.
Logicity's Take
Frequently Asked Questions
How much are Big Tech companies spending on AI in 2026?
Microsoft, Alphabet, Meta, and Amazon have combined AI capital expenditure guidance of $725 billion for 2026, more than three times the inflation-adjusted cost of the Manhattan Project.
Which company had the fastest AI revenue growth in Q1 2026?
Google Cloud grew 63% year-over-year to $20 billion in quarterly revenue, its highest growth rate ever, driven by generative AI demand from enterprise customers.
Is Microsoft's AI business profitable?
Microsoft's AI revenue reached a $37 billion annualized run-rate, up 123% from last year. CFO Amy Hood noted the company is capacity-constrained, with demand exceeding available supply through the end of fiscal 2026.
How is Meta making money from AI?
Meta has built a $10 billion annual revenue run-rate from AI-generated video ads through its Advantage+ advertising system, though investors remain skeptical because AI gains are embedded in advertising rather than reported separately.
Why isn't Apple spending as much on AI infrastructure?
Apple doesn't operate hyperscale cloud infrastructure or sell AI services to enterprises. Its AI strategy runs through device features and partnerships like Apple Intelligence with OpenAI, making it an AI consumer rather than an infrastructure provider.
Need Help Implementing This?
Source: Stratechery by Ben Thompson
Post-Earnings Update: Apple AI Hardware Expansion
Apple has launched the MacBook Neo, an affordable laptop featuring the A18 Pro chipset and a 16-core Neural Engine designed to run Apple Intelligence on macOS 26. This development, coupled with its promotion in the Amazon Summer Sale, provides concrete evidence of how the companies are translating their record AI spending into consumer-facing products following their Q1 2026 earnings.
Manaal Khan
Tech & Innovation Writer
Produced with AI assistance and reviewed by the Logicity editorial team. Learn more in our Editorial Policy.
