Who Owns the Future? The Race to Control AI Compute

Every AI system, the ones writing school essays, screening job applications, reading medical scans, runs on physical hardware somewhere. Warehouses full of specialised chips, humming in the desert or on the edge of a power grid. Whoever controls those chips controls AI. And right now, that’s about four companies.

In 2026, they’re spending more to build this infrastructure than any nation spends on defence. This is who they are, how the money flows, and why it matters to everyone.

The compute leaderboard

AI runs on raw computing muscle, massive clusters of specialised chips. Google, Microsoft, Meta, and Amazon each operate fleets worth hundreds of thousands of processors. Behind them, xAI and OpenAI are scaling fast.

Source: Epoch AI Computing Capacity Analysis, company earnings calls, SemiAnalysis estimates. 2025 projections are analyst estimates, not company-disclosed.

The United States hosts roughly 75% of the world’s AI compute. China holds about 15%. Everyone else shares the scraps. And the total is doubling every seven months.

The investment arms race

Follow the money. In 2024, the five biggest cloud companies spent a combined $256 billion on infrastructure. In 2025, that hit $443 billion. In 2026, company guidance puts the total at $645 billion, the largest infrastructure buildout in human history.

Source: Company Q4 2025 earnings reports (Amazon, Alphabet, Meta, Microsoft). 2026 figures are company guidance midpoints.

Google alone plans to spend $175-185 billion this year, nearly doubling last year’s $91.4 billion. The Stargate project, a joint venture between OpenAI, SoftBank, and Oracle, aims to deploy $500 billion over four years. To finance all this, these companies took on $121 billion in new debt in 2025 alone.

These aren’t research budgets. This is concrete, copper, and cooling systems. Physical bets that AI will be worth it.

The supply chain that holds it all together

All of this power flows through an astonishingly narrow pipeline. One company, NVIDIA, designs 94% of AI chips. One manufacturer, TSMC in Taiwan, fabricates over 90% of the advanced chips those designs require. One memory supplier, SK Hynix in South Korea, provides 62% of the specialist memory that makes them work.

Three companies. Two countries. Zero margin for disruption.

Source: Mercury Research (GPU share), TrendForce (TSMC foundry share), Astute Group (HBM memory share), 2024-2025.

Every major tech company is racing to design its own chips, Google, Amazon, Microsoft, but NVIDIA keeps raising the bar. Meanwhile, Intel has effectively left the race, and Samsung has lost half its memory market share in a single year.

What comes next

If current trajectories hold, global AI compute will grow 10x by late 2027. Spending will approach $1 trillion per year. And the power required to run it all is already straining national electricity grids, Ireland’s data centres may consume 32% of the country’s electricity by 2026.

Source: IEEE ComSoc (aggregate capex), NVIDIA Investor Relations (quarterly revenue). Q4 FY26 is guidance.

The decisions being made right now, by a small number of companies, in a small number of countries, will determine who builds the AI systems that reshape how people work, get medical care, and are governed. The computing power is the foundation. Everything else is built on top.