📊 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 energy grid and renewable buildout to deploy AI data centers at gigawatt scales, bypassing US transmission constraints. The US remains dominant in chips and models but faces structural limits at the power delivery layer.
China’s centralised energy infrastructure and extensive renewable buildout are enabling the deployment of gigawatt-scale AI data centers, contrasting with the US’s fragmented grid that constrains its infrastructure growth. This structural difference may influence global AI leadership in the coming years.
Current frontier AI data centers require 100 megawatts to start and up to 2 gigawatts at full buildout, with the largest US projects targeting capacities of 12 GW. The US relies on behind-the-meter agreements, off-grid gas turbines, and regulatory arbitrage to reach these scales, facing significant transmission bottlenecks.
China, by contrast, has routed eastern AI demand to western renewable hubs through 45 ultra-high-voltage (UHV) transmission projects spanning over 40,000 kilometers, with a capacity of 340 GW. In 2025, China added over 430 GW of wind and solar, surpassing US renewable additions by a factor of eight, and now has a total installed capacity of 3.89 TW.
Although Chinese AI chips, such as Huawei’s Ascend 910C, perform at roughly 60% of NVIDIA’s H100 inference levels and lack native FP8/FP4 support, their deployment across China’s extensive and centralized power infrastructure effectively compensates for raw chip performance. This system-level asymmetry allows China to substitute raw power capacity for chip performance, a strategic choice rooted in structural differences between the US and China.
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 Power Infrastructure for AI Dominance
This analysis suggests that AI leadership may increasingly depend on the ability to deploy and operate large-scale data centers, which are constrained by physical power delivery infrastructure. China’s centralized planning and renewable energy strategy provide a structural advantage, potentially enabling faster and larger AI deployments despite lower chip performance. The US’s fragmented grid and regulatory hurdles could impose a ceiling on its AI infrastructure growth, influencing global AI competitiveness.
gigawatt-scale AI data center power supplies
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
US versus China: Divergent AI Infrastructure Strategies
The US leads in chip design, AI models, and software applications but faces physical limitations at the power delivery layer. Its data centers, often built on off-grid or regulated grids, require complex, costly, and time-consuming permitting processes, limiting their scale.
China’s approach leverages centralised planning, extensive renewable energy projects, and ultra-high-voltage transmission to bypass these constraints. The country’s renewable capacity grew by approximately eight times US additions in 2025, facilitating gigawatt-scale AI data centers that operate at the system level more efficiently than US facilities.
While Chinese chips lag in raw performance, their deployment across China’s vast, centralized, renewable-powered grid allows for a different metric of AI capability—one that emphasizes raw power throughput over chip-level performance.
“The gigawatt-scale capacity requirements of frontier AI deployments are now fundamentally different from previous megawatt-scale data centers, relying on power generation and transmission at an industrial scale.”
— Thorsten Meyer
G1-10 Ultra High Voltage Diode 10KV 1000mA, Unidirectional Rectifier Diode for X-Ray Equipment, Industrial Power Supply, HV Generators (100-Pack)
Superior High Voltage Handling: Engineered as a dedicated High Voltage Diode, the G1-10 delivers unparalleled stability under extreme…
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Unresolved Questions on Infrastructure and Policy Impact
It remains unclear whether US efficiency gains in chips, racks, and models will close the gap at the power layer or whether structural constraints will impose a sustained ceiling on US AI infrastructure growth. The impact of potential regulatory reforms or new grid investments is still uncertain.
The BESS Book: A Cell to Grid Guide to Utility-Scale Battery Energy Storage Systems
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Future Developments in AI Infrastructure Strategies
Over the next 24 months, monitoring US policy reforms, renewable buildout, and grid expansion efforts will be critical to assess whether the US can overcome structural constraints. Meanwhile, China’s continued infrastructure investment and deployment will further test the competitive dynamics of AI leadership.
Jackery Solar Generator 1000 v2 with 200W Solar Panel,1070Wh Portable Power Station LiFePO4 Battery,1500W AC/100W USB-C Output, 1Hr Fast Charge for Outdoor,Off-Grid Living,RV,Emergency
Powerful yet Compact: Boasting a 1,500W AC output and a 3,000W surge peak, the Solar Generator 1000 V2…
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
Why does power infrastructure matter more than chip performance in AI deployment?
Because AI data centers at frontier scale require gigawatt-level power throughput, and the ability to deliver that power reliably and affordably is a key bottleneck. Chip performance alone does not determine the capacity to run large-scale AI systems.
How does China’s renewable energy strategy influence its AI infrastructure?
China’s extensive renewable buildout and centralized transmission enable large-scale, gigawatt-capacity data centers that are less constrained by regulatory or transmission bottlenecks, giving it a structural advantage at the infrastructure layer.
Could the US overcome its infrastructure constraints through policy or technological improvements?
It is uncertain. While efficiency gains and regulatory reforms could help, the current fragmentation and regulatory complexity pose significant hurdles that may limit the pace and scale of US AI infrastructure expansion.
Does chip performance still matter for AI leadership?
Yes, but at the system level, power throughput and infrastructure capacity are becoming equally or more important. Lower chip performance can be compensated by deploying more chips across larger, centralized power systems, as China is doing.
What are the potential risks if China’s infrastructure advantage persists?
If China maintains its infrastructure lead, it could accelerate its AI deployment and capabilities, challenging US dominance. However, this depends on technological developments, policy choices, and global energy trends.
Source: ThorstenMeyerAI.com
