AI Data Center Capacity Doubles Every 7 Months as Hyperscaler Spending Reaches $725 Billion

Epoch AI analysis reveals computing capacity at the largest AI data centers has been doubling every seven months since August 2024, while hyperscaler infrastructure spending surges 77% year-over-year to $725 billion.

The largest AI data centers are scaling at a pace that outstrips even the most ambitious forecasts. According to new analysis from Epoch AI, the record for computing capacity in a single data center has doubled every seven months since SpaceXAI's Colossus 1 launched in August 2024 — a growth rate of 3.3× per year.

Colossus 2, operated by SpaceXAI in Memphis, Tennessee, currently holds the crown at 1.1 million H100-equivalent chips. It is followed by Microsoft Fairwater Atlanta (769,000), Meta Prometheus (763,000), and the Anthropic-Amazon facility in New Carlisle, Indiana (686,000). OpenAI's Stargate campus in Abilene, Texas, rounds out the top tier.

Epoch AI, which tracks these facilities using satellite imagery, construction permits, and public disclosures, estimates that 5.2 million H100-equivalent chips were operational globally as of March 2026 — out of 20.2 million shipped. The research group covers roughly 26% of global AI compute, focused primarily on the largest facilities in the United States.

The trendline points to a plateau through mid-2027, as no single facility is projected to surpass Colossus 2 until the second half of that year. But the lull may be brief: QTS Cedar Rapids and Meta Hyperion are both expected to break the record in late 2027 or early 2028.

The infrastructure bill underpinning this expansion is staggering. A separate analysis by The ByteDive finds that Amazon, Microsoft, Alphabet, and Meta will collectively spend $725 billion on AI infrastructure in 2026 — a 77% increase over 2025. Goldman Sachs projects cumulative spending could reach $7.6 trillion through 2031, a figure that would require over $1 trillion in annual AI revenue to justify.

NVIDIA remains the dominant supplier, with $215.9 billion in revenue, though AMD's MI400 and custom silicon from Google (TPU v6) and Amazon (Trainium2) are beginning to diversify the hardware landscape.

For now, the doubling continues. The question is whether AI revenue can catch up to the capital pouring into its physical foundation.

Sources: Epoch AI — Largest AI Data Center: Doubling Every 7 Months — The ByteDive — AI Infrastructure Value Chain 2026: $725B vs $1T Gap