---
id: "claim-us-compute-dominance"
type: "claim"
source_timestamps: ["§ The Leaders."]
tags: ["hardware", "infrastructure", "us"]
related: ["concept-the-leaders", "entity-trg-datacenters", "prereq-ai-compute-metrics", "claim-us-china-ai-gap-closed"]
confidence: "high"
testable: true
sources: ["futures"]
sourceVaultSlug: "hbr-seg-futures"
originDay: 2
articleStem: "hbr-foci-75-fragmenting-digital-economy"
sourceUrl: "https://hbr.org/2026/06/what-a-fragmenting-digital-economy-means-for-global-competition"
sourceTitle: "What a Fragmenting Digital Economy Means for Global Competition"
---
# U.S. Holds Half of Global AI Compute

**Claim:** According to [[entity-trg-datacenters]], the United States possesses an estimated AI compute capacity of **39.7 million petaflops** — approximately **half of the world's total** processing power dedicated to AI — a massive hardware advantage over China's estimated **400,000 petaflops**.

Understanding why this disparity matters requires [[prereq-ai-compute-metrics]] (a petaflop = one quadrillion floating-point operations per second). The gap makes China's algorithmic-efficiency counter-strategy in [[claim-us-china-ai-gap-closed]] all the more notable.

> **Enrichment — directionally plausible, precision cautious:** U.S. dominance in data centers, GPUs, and cloud AI capacity, and a large gap with China, are widely discussed (AI Index, chip-export-control literature). But TRG Datacenters' exact figures (39.7M vs. 400k petaflops) and the "half of global AI compute" claim are *not* broadly corroborated by independent measurement sources — treat the precise numbers cautiously.