📊 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 a centralized, renewable-powered infrastructure to deploy AI at gigawatt scales, bypassing US grid constraints. The US leads in chips but faces structural limits at the power delivery layer, potentially impacting future AI competitiveness.
China’s centralized infrastructure and extensive renewable energy buildout are enabling it to deploy AI data centers at gigawatt-scale capacities, circumventing the US’s grid constraints and potentially shifting the global AI power balance.
Current frontier AI data centers require 100 MW to 2 GW of power, with the largest projects reaching up to 12 GW. The US relies on a fragmented grid system, off-grid gas turbines, and regulatory arbitrage to meet these demands, leading to long interconnection queues and permitting challenges.
In contrast, China employs a centralized, top-down approach, routing eastern AI demand through the Western Data and Compute initiative, which connects renewable hubs via over 40,000 km of ultra-high-voltage (UHV) transmission lines capable of 340 GW capacity. In 2025, China added over 430 GW of wind and solar, more than eight times the US increase, supporting its AI infrastructure.
Chinese AI chips, such as Huawei’s Ascend 910C, perform at about 60% of US chips like the NVIDIA H100, but the system-level capacity—powered by abundant renewables and extensive transmission—compensates for lower per-chip performance. This structural difference means China substitutes raw power availability for chip-level performance, a strategic choice rooted in its centralized planning and infrastructure scale.
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 on AI Global Leadership
This structural divergence could determine the future of AI leadership. While the US maintains dominance in chip innovation and AI software, China’s ability to scale AI infrastructure through renewable energy and extensive transmission may allow it to deploy AI at larger scales more rapidly. The shift from performance-per-chip to power throughput as a key metric challenges traditional assumptions about AI capacity and competitiveness.
Understanding this dynamic is vital for policymakers, industry leaders, and investors, as the next two years could see a realignment of global AI power based on infrastructure capabilities rather than chip performance alone.
Protection Technologies of Ultra-High-Voltage AC Transmission Systems
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
US and China Approaches to AI Infrastructure Development
US AI infrastructure development has been constrained by regulatory, permitting, and grid limitations, leading to reliance on off-grid solutions and complex interconnection processes. Major US projects like Meta Hyperion and OpenAI Stargate are approaching 5 GW capacities but face bottlenecks due to grid congestion and regulatory delays.
China, by contrast, has adopted a centralized model, integrating renewable energy expansion with ultra-high-voltage transmission to connect remote renewable hubs directly to AI data centers. This approach allows for gigawatt-scale deployments without the same regulatory hurdles faced in the US.
While Chinese chips lag in raw performance, the scale and efficiency of their power infrastructure enable them to deploy AI systems at a system level that rivals or surpasses US capabilities, especially as AI models grow larger and more energy-intensive.
“The US dominates AI hardware and software, but at the physical power layer, China’s centralized, renewable-powered infrastructure provides a structural advantage that could reshape global AI deployment dynamics.”
— Thorsten Meyer
gigawatt-scale data center power supplies
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Unresolved Questions About Future Infrastructure and Policy
It remains unclear whether the US can overcome its grid and permitting constraints through policy reforms or technological efficiency gains, and whether these efforts will suffice to close the gigawatt gap. Additionally, the long-term impact of China’s infrastructure-led approach on global AI leadership is still developing, with potential shifts depending on geopolitical and technological factors.
Advanced Concepts for Renewable Energy Supply of Data Centres
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Next Steps in Monitoring AI Infrastructure Developments
Over the coming 24 months, observers will watch US policy efforts to reform permitting and grid expansion, alongside technological advances in chip efficiency. Simultaneously, China’s ongoing renewable expansion and infrastructure investments will be key indicators of whether its system-level approach can sustain its advantage or if US innovations can bridge the power gap.
Further analysis of how these structural differences influence AI deployment speed, model scaling, and international competitiveness will be critical in assessing future leadership in AI technology.
large-scale AI data center UPS systems
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 for AI scaling?
Because large AI models require vast amounts of energy, the capacity to supply power at gigawatt scales becomes a limiting factor. Infrastructure that can deliver abundant, reliable power enables deployment at larger scales, even if individual chips are less powerful.
Can the US overcome its grid and permitting constraints to compete with China’s infrastructure advantage?
It is uncertain. Policy reforms, technological improvements, and regulatory changes could help, but the scale and complexity of US grid constraints pose significant challenges that may take years to resolve.
Will China’s reliance on lower-performance chips limit its AI capabilities?
Not necessarily. China’s infrastructure approach compensates for chip performance gaps by enabling larger-scale deployment and energy efficiency, which can support comparable or even superior AI system capacity at the system level.
How might these infrastructure differences impact global AI leadership?
If China’s infrastructure approach proves more scalable and cost-effective, it could shift AI leadership toward regions with centralized, renewable-powered grids, challenging the US dominance based on chip innovation alone.
Source: ThorstenMeyerAI.com
