Deep Dive: Why China's East-Data-West-Computing Falls Short
ChinaTalk · rss · 2026-07-10
The article points out that China's "East-Data-West-Computing" project hasn't actually transferred massive computing power to the west in practice; it's mostly an overflow from eastern core cities to their surrounding suburbs.
Current Computing Distribution
Data shows that China's computing power remains highly concentrated in eastern coastal regions (like Hebei, Guangdong, Jiangsu) and mega-city clusters. Although remote western provinces have data centers, they mainly serve the inference needs of sparse local populations and haven't formed the expected national-level computing hubs.
Economics and Supply Chain Constraints
- Electricity isn't the main cost: Despite abundant renewable energy in the west, electricity only accounts for about 5% of the total three-year cost of large AI data centers. The core expenses are chip procurement and facility construction.
- Talent and ops shortages: Data centers require dense hardware, network, and cooling talent, but 87.5% of China's relevant talent is concentrated in the North, East, and South, making it hard for the west to support advanced computing operations.
- Hardware bottlenecks: Semiconductor export controls remain a core obstacle for China to achieve ultra-large-scale computing deployments.
Policy Reflection
The author argues that the "East-Data-West-Computing" slogan is disconnected from the real geographic and economic logic of AI-era data centers. Blindly pushing less-developed western regions to take on debt for data center construction not only faces extremely high latency but could also make them the biggest losers in terms of policy and development models.
More from Infra
- SkyPilot emerges from stealth with over $20M to tackle fragmented AI compute — skypilot_org · 2026-07-22
- Why a 1GW Chinese AI data center may be plausible after all — teortaxesTex · 2026-07-22
- China’s AI arms race is increasingly defined by chips, data centers, and open models — BenBajarin · 2026-07-22
- Agent search bottlenecks are now about variance, not raw latency — rohanpaul_ai · 2026-07-22
- Gavin Baker argues Nvidia may be one of open source AI’s biggest supporters — GavinSBaker · 2026-07-22
- AI Power Demand Exposes US Energy Gap, Urging Shift from Scarcity to Abundance — bradneuberg · 2026-07-22