📊 Full opportunity report: The gigawatt gap. Why China is structurally positioned for AI power and the US is engineering around its grid. on ThorstenMeyerAI.com — validation score, market gap, and execution plan.
TL;DR
China is leveraging its centralized planning and renewable energy infrastructure to deploy AI data centers at gigawatt capacities, while the US faces constraints at the power delivery layer. This structural difference could reshape global AI leadership.
China is deploying AI data centers at gigawatt-scale capacities through centralized planning and extensive renewable energy infrastructure, challenging the US’s dominance at this critical physical layer of AI infrastructure development.
Recent analysis by Thorsten Meyer explains that while the US leads in chips, models, and AI applications, it faces significant constraints at the power delivery layer due to regulatory, permitting, and grid fragmentation issues. In contrast, China’s approach leverages a centralized infrastructure system, with the NDRC’s Eastern Data Western Compute initiative routing demand across 45 ultra-high-voltage transmission projects, enabling the transfer of over 340 GW of capacity. In 2025, China added approximately 430 GW of wind and solar power—eight times the US’s renewable additions—pushing total renewable capacity to over 1.8 TW and overall capacity to nearly 3.89 TW.
Chinese AI chips, such as Huawei’s Ascend 910C, perform at roughly 60% of US chips like the NVIDIA H100, but the Chinese system compensates by substituting raw power for chip performance. This structural advantage stems from China’s centralized planning, extensive renewable infrastructure, and vast transmission network, which collectively enable deploying less efficient chips across more power capacity without the same regulatory constraints faced by the US.
The gigawatt gap.
Why China is structurally
positioned for AI power
and the US is engineering
around its grid.
power capacity end 2025
5-year average wait
45 projects · 340 GW capacity
vs. H100 · compensated by watts
interconnection queue
installed capacity
built by end-2024
on-site generation
DY 2024-25 → 2026-27
solar additions 2025
generation capacity
installed base
of capacity
add ratio
2025 alone
capacity end 2025
installed capacity
of capacity
Low watts
grid + transmission capacity
More watts
chip performance / FP precision
The US has perf-per-watt advantage. China has watts-without-bound advantage. These are asymmetric substitutes — not the same axis. When the perf-per-watt side is bounded by grid capacity and the watts-without-bound side is bounded by chip performance, the binding constraint differs.Thorsten Meyer · The Gigawatt Gap · Energy & Infrastructure 01
Implications of the Gigawatt Power Gap in AI Infrastructure
This structural difference in infrastructure could redefine global AI leadership. China’s ability to deploy AI at gigawatt scales, supported by renewable energy and centralized planning, may offset the US’s technological edge in chip performance. The outcome will influence who controls the physical layer of AI deployment, with potential long-term impacts on innovation, cost, and geopolitical influence.
gigawatt-scale data center power supplies
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
US and China Approaches to AI Infrastructure Development
The US has built an AI infrastructure stack focused on chips, models, and software, but faces bottlenecks at the physical power delivery layer due to grid constraints and regulatory complexity. Major US data centers now require 100 MW to start and up to 2 GW at full buildout, with a backlog of interconnection requests exceeding 2,300 GW. Meanwhile, China’s strategy involves large-scale renewable generation combined with ultra-high-voltage transmission to supply AI data centers across vast distances, effectively bypassing regulatory and transmission bottlenecks. This approach is enabled by centralized planning and rapid renewable deployment, with China adding about 8 times more renewable capacity in 2025 than the US.
“The US infrastructure is constrained at the power delivery layer, while China’s centralized system and renewable buildout enable gigawatt-scale deployment that bypasses many US bottlenecks.”
— Thorsten Meyer
Extruded Cables for High-Voltage Direct-Current Transmission: Advances in Research and Development (IEEE Press Series on Power and Energy Systems)
Used Book in Good Condition
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Unclear Impact of Efficiency Gains and Policy Changes
It remains uncertain whether US efforts to improve chip efficiency and reform infrastructure permitting will close the gigawatt gap or whether China’s centralized approach will maintain its advantage. The long-term effects of these structural differences on global AI leadership are still developing and depend on future policy, technological, and geopolitical developments.
The Electric Hunger of AI: How Data Centers, Power Grids, Energy Wars Will Shape the Future of Artificial Intelligence – The Hidden Energy Crisis Behind Artificial Intelligence
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Potential Developments in US and Chinese AI Infrastructure Strategies
Over the next 24 months, attention will focus on whether the US can implement regulatory reforms, improve infrastructure permitting, or develop new energy solutions to overcome power bottlenecks. Simultaneously, China’s continued renewable expansion and infrastructure investments will be monitored to assess whether their centralized approach sustains its competitive edge. The outcome will shape the global AI deployment landscape and influence future technological and geopolitical balances.
How AI Uses Our Water: When Machines Get Thirst: Cooling Systems, Data Centres, and the Infrastructure Behind Artificial Intelligence
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
Why does the US face constraints at the power layer for AI data centers?
The US’s fragmented grid system, regulatory hurdles, and permitting delays create bottlenecks that limit the scale and speed of deploying gigawatt-scale AI data centers.
How is China able to deploy AI infrastructure at such large scales?
China’s centralized planning, extensive renewable energy buildout, and ultra-high-voltage transmission network allow for large-scale power transfer and deployment of AI data centers across vast distances.
Does chip performance still matter in AI deployment?
Yes, but at the system level, the total power throughput becomes a critical constraint, meaning that raw chip performance is less decisive than the ability to supply sufficient power at scale.
Could the US close the gigawatt gap through efficiency improvements?
It is uncertain; efficiency gains in chips and infrastructure reform could help, but structural constraints may limit the extent of closing the gap compared to China’s centralized approach.
What are the long-term implications of these structural differences?
The outcome will influence which country leads in AI deployment, innovation, and geopolitical influence, depending on whether power infrastructure or chip performance becomes the dominant limiting factor.
Source: ThorstenMeyerAI.com
