The numbers almost defy comprehension. The five largest hyperscale technology companies—Amazon, Microsoft, Google, Meta, and Apple—are projected to spend $602 billion on capital expenditures in 2026, according to CreditSights. That's a 36% increase from the approximately $443 billion they're on track to spend in 2025, and approximately 75% of it will go directly into artificial intelligence infrastructure.

The Spending Breakdown

Each of the major hyperscalers has dramatically increased its infrastructure investment as AI capabilities become central to competitive positioning:

  • Amazon: $125 billion in 2025 capex, with guidance suggesting further increases in 2026
  • Alphabet (Google): $92 billion at the midpoint of 2025 guidance, up from $85 billion earlier in the year
  • Microsoft: $80 billion in 2025, with quarterly spending running ahead of analyst expectations
  • Meta: Aggressive expansion of AI infrastructure to support generative AI features

Goldman Sachs projects that these companies plus Oracle will collectively spend $533 billion on AI-related infrastructure in 2026, while CreditSights' broader $602 billion estimate includes all capital expenditures.

Where the Money Is Going

The spending is concentrated in a few key areas:

Data Centers

AI workloads require specialized data centers with unique power, cooling, and networking requirements. Microsoft is building what CEO Satya Nadella calls "the world's most powerful" AI data center in southeast Wisconsin—a $7 billion facility that will house hundreds of thousands of Nvidia chips when it comes online in early 2026.

Amazon has transformed 1,200 acres of Indiana farmland into Project Rainier, an $11 billion facility running entirely on custom silicon, built exclusively to train AI models for Anthropic.

Custom Chips

While Nvidia dominates AI chip supply, hyperscalers are increasingly developing custom silicon to reduce costs and optimize for their specific workloads:

  • Amazon: Trainium and Inferentia chips for training and inference
  • Google: Tensor Processing Units (TPUs) now in their fifth generation
  • Microsoft: Maia 100 AI accelerators
  • Meta: MTIA (Meta Training and Inference Accelerator)

Power Infrastructure

AI data centers consume enormous amounts of electricity. Meta recently announced plans to secure 6.6 gigawatts of nuclear power capacity to support its AI ambitions—enough to power millions of homes.

"We've been short on computing power now for many quarters. I thought we were going to catch up. We are not. Demand is increasing."

— Amy Hood, Microsoft CFO

The Debt Dimension

Financing this spending spree has required unprecedented corporate borrowing. According to Bank of America, hyperscalers have added $121 billion in new debt in 2025—more than four times the average annual issuance over the previous five years.

By 2026, several major tech companies are projected to generate negative free cash flow after shareholder returns:

  • Meta: Negative free cash flow projected after dividends and buybacks
  • Microsoft: Negative free cash flow projected after shareholder returns
  • Alphabet: Expected to break even

This represents a dramatic shift for companies that have historically been cash-generating machines. The bet is that AI capabilities will generate returns that justify the investment—but the payoff timeline remains uncertain.

Investment Implications

The AI spending bonanza creates both opportunities and risks for investors:

Beneficiaries

  • Nvidia (NVDA): Remains the dominant supplier of AI training chips, with demand consistently outstripping supply
  • TSMC (TSM): Manufactures chips for virtually all major AI players; just raised capex guidance
  • ASML (ASML): Its EUV lithography machines are essential for advanced chip production
  • Utilities: Data center power demand is creating unprecedented opportunities for electricity providers
  • Industrial REITs: Data center real estate commands premium valuations

Potential Risks

  • Overinvestment: If AI revenue fails to materialize as expected, hyperscalers could face write-downs
  • Rising rates: Higher borrowing costs could pressure returns on debt-financed investments
  • Regulatory risk: AI regulation could limit applications and reduce demand growth
  • Competition: Custom chips from hyperscalers could eventually pressure Nvidia's market share

The Bull and Bear Case

The Bull Case

AI proponents argue that we're in the early stages of a technological transformation comparable to the internet or electricity. From this perspective, $602 billion in annual spending is appropriate—even conservative—given the potential for AI to reshape entire industries.

Key supporting arguments:

  • Enterprise AI adoption is accelerating across all sectors
  • AI agent technology promises to automate complex workflows
  • Productivity gains could justify premium pricing for AI services
  • First-mover advantages in AI could prove durable

The Bear Case

Skeptics point to historical precedents of technology overinvestment, from the telecom fiber glut of the early 2000s to the cryptocurrency mining boom of 2021. Warning signs include:

  • Diminishing returns on AI model improvements
  • Limited monetization of generative AI features to date
  • Rising costs potentially squeezing margins
  • Competition potentially commoditizing AI capabilities

What to Watch

For investors monitoring the AI spending cycle, key indicators include:

  • Hyperscaler earnings calls: Listen for commentary on AI revenue and return on AI investments
  • Nvidia's backlog: Currently estimated at $500 billion, any deceleration would be significant
  • Cloud revenue growth: AI-driven acceleration in AWS, Azure, and Google Cloud growth rates
  • Enterprise AI adoption surveys: Track whether corporate customers are increasing AI spending

The $602 billion question is whether AI's transformative potential will justify the largest corporate infrastructure investment in history. The hyperscalers are betting their balance sheets that the answer is yes. Investors must decide whether to join them.