AI Compute Grows 5x in 7 Quarters: 2GW to 10GW Deployed

Total deployed AI compute has grown from roughly 2 gigawatts to over 10 gigawatts in just seven quarters—a fivefold increase. NVIDIA dominates every period, accounting for approximately 70-75 percent of cumulative capacity.

The Concentration Reality

Google’s TPUs represent the second-largest share, with AMD, Huawei, and Amazon’s custom silicon contributing thin slivers at the base. Despite years of competitor investment and billions spent on alternatives, NVIDIA’s market share remains overwhelming.

Provider	Share	Trend
NVIDIA	~70-75%	Dominant across all periods
Google TPU	~15-20%	Growing but distant second
AMD	~5%	Gaining slowly
Others (Huawei, Amazon)	~5%	Thin slivers

The 5x Growth Story

The scale of AI infrastructure buildout is remarkable:

• Q1 2024: ~2 gigawatts deployed
• Q4 2025: 10+ gigawatts deployed
• Growth rate: 5x in under two years

This represents one of the fastest infrastructure buildouts in technology history—comparable to the early days of cloud computing but compressed into a shorter timeframe.

The Strategic Question

Cumulative AI chip sales have grown fivefold, but concentration has not decreased. The question is whether Google’s TPU growth and eventual Broadcom/AMD alternatives can create meaningful diversification—or whether NVIDIA’s moat is structural. NVIDIA’s advantages compound:

• CUDA ecosystem: Software lock-in across millions of developers
• Full-stack integration: Hardware + software + networking
• Scale economics: Largest R&D budget, fastest iteration

What This Means

The AI infrastructure supercycle continues, but it’s largely a NVIDIA supercycle. Competitors are growing in absolute terms while losing share relatively. For investors and enterprises, the vertical integration moat appears durable. Source: Industry Analysis