🔥 WHAT HAPPENED

For the first time ever, all four hyperscalers — Alphabet, Amazon, Meta, and Microsoft — reported earnings on the same day. And the headline number is genuinely jaw-dropping: more than $650 billion in combined AI capital expenditure planned for 2026 alone.

That's larger than the GDP of Switzerland. Or Sweden. Or about 150 other countries.

Let's put this in perspective. Alphabet alone raised its 2026 capex forecast to $180–190 billion (up $5 billion from last quarter), with CFO Anat Ashkenazi explicitly stating 2027 will "significantly increase." Microsoft spent $31.9 billion in Q1 alone (up 49% YoY), on pace for ~$190 billion total. Amazon rivals that at roughly $200 billion across 2026. Even Meta — whose stock got punished for it — raised its range to $125–145 billion.

These are not normal numbers. These are "we've decided the future is AI and we're buying it" numbers.

🧠 WHY THIS MATTERS

This is the moment the AI bubble debate gets a conclusive answer — or at least, the market delivers its verdict.

The bear case says this capex is an insane, unprecedented bet that will end in write-offs and crushed returns. The bull case says it's working: Google Cloud hit $20 billion quarterly revenue for the first time (up 63% YoY), and for the first time Google says AI workloads drove the majority of that growth. AWS posted $37.6 billion (up 28%, fastest growth in 15 quarters). Microsoft's Azure grew 40%.

The market effectively voted: Alphabet and Amazon shares rose. Meta fell 6%. The difference? Meta's spending has no visible ROI story yet. A reminder that investors will reward AI capex — but only if you show them the revenue.

📊 DEEP DIVE

Let's break down the numbers:

  • Revenue: $109.9B (beat $107.2B estimates)
  • Google Cloud: $20B — first time over the milestone, up 63% YoY
  • Capex 2026: $180–190B, raised from $175–185B
  • CFO says 2027 capex will "significantly increase"
  • AI workloads now the primary driver of cloud growth for the first time

  • Azure growth: 40%
  • Q1 capex: $31.9B (up 49% YoY from $21.4B)
  • Annualized AI revenue run rate: ~$37B
  • On track for ~$190B total 2026 capex
  • CEO Satya Nadella: "Every workload is being reinvented with AI"

  • AWS revenue: $37.6B (beat $36.64B estimate)
  • AWS growth: 28% — fastest in 15 quarters
  • 2026 total capex: ~$200B
  • AWS CEO Matt Garman says AI demand is "unprecedented"
  • Amazon's chip business (Trainium/Inferentia) topped $20B revenue run rate

  • Capex Q1: $19.84B (below $27.57B estimate — timing, not reduction)
  • Full-year range raised to $125–145B (from $115–135B)
  • Stock fell 6% after hours — investors want ROI, not just spending
  • Reality Labs losses still around $19B

Combined total: ~$650B+ in 2026 AI infrastructure capex

⚠️ THE CATCH

There are real risks here:

Meta's Roadmap Problem. Mark Zuckerberg is spending $125B+ with no clear AI revenue story beyond ads. The stock drop reflects real skepticism. If Meta's AI investments don't show returns in 2026, expect activist pressure.

Pentagon Paradox. Alphabet just signed a classified deal with the Pentagon for "any lawful government purpose" — two months after the DOD dropped Anthropic as a supply chain risk. Cameron Stanley, the Pentagon's AI chief, said "reliance on one model is never a good thing." Google also expanded to classified networks, putting it alongside OpenAI and xAI. The irony? Anthropic got blacklisted partly because of its own ethical stance on military AI.

Iran War Supply Chain Risk. CNBC explicitly called out that surging oil prices and supply chain disruptions from the Iran conflict could inflate infrastructure costs. The hyperscalers are placing massive bets on hardware supply chains that face real geopolitical headwinds.

OpenAI Just Blew Up Microsoft Exclusivity. Two days ago, OpenAI ended its exclusive deal with Microsoft and landed on AWS Bedrock in a massive deal. OpenAI can now run on AWS, Google Cloud, and Azure simultaneously. That's great for competition, but it fundamentally changes the Microsoft-OpenAI relationship that drove much of the early AI narrative.

EU Regulation Collapsed. After 12 hours of trilogue negotiations on Tuesday, EU member states and the European Parliament failed to reach a deal on the AI Act. Talks resume in May — but with the August 2026 compliance deadline looming, uncertainty is the only certainty.

🎯 WHAT HAPPENS NEXT

2027 is already spoken for. Alphabet's CFO explicitly guided higher capex next year. This isn't a one-year sprint — it's a multi-year infrastructure buildout that will reshape cloud computing, energy grids, and the semiconductor industry.

The Anthropic-OpenAI arms race just got absurd. Anthropic is weighing a $50 billion raise at $850–900 billion valuation — leapfrogging OpenAI. The Claude maker has multiple preemptive offers. If that round closes, Anthropic becomes the most valuable private AI company in history, and it's only 5 years old.

Short-term pain, long-term gain. The earnings calls told a clear story: the companies showing AI revenue (Alphabet, Amazon, Microsoft) got rewarded. Meta didn't, and its stock suffered. Expect more pressure on every company to connect capex to cash flow.

Regulation will be chaotic. The EU talks collapsing means Europe enters a compliance deadline with no clarity. Meanwhile the US Pentagon is going all-in on multiple vendors. The regulatory landscape is fragmented and getting more so.

🧩 BIGGER PICTURE

This isn't just about earnings. It's the clearest signal yet that the AI industry has entered its infrastructure phase — the equivalent of building the interstate highway system before the cars arrive.

The hyperscalers are spending like there's no tomorrow because they believe AI demand is structurally unlimited. The numbers back it up for now: cloud growth accelerating at historic rates, AI workloads becoming the primary growth driver, and every enterprise wanting their own custom model.

But $650 billion is real money with real opportunity cost. The companies that can connect this spending to actual product revenue will thrive. The ones that can't will face angry shareholders and activist campaigns.

For the rest of us: expect faster AI capabilities, more powerful models, and — eventually — lower prices as this infrastructure gets built out. The compute arms race is just getting started.