📊 Full opportunity report: The Challenge Of Providing Enough Energy For AI on ThorstenMeyerAI.com — validation score, market gap, and execution plan.
TL;DR
The expansion of AI infrastructure is constrained by physical power capacity, not funding. The US and China face different energy challenges that could impact AI development timelines.
Global data-center capacity is rapidly increasing, but power grid limitations are emerging as the primary bottleneck for AI infrastructure expansion, according to recent analysis. This shift from chip scarcity to energy capacity constraints is reshaping the AI race and infrastructure planning.
Data-center capacity worldwide is projected to reach approximately 290 GW by 2030, up from 132 GW in 2026. However, power grid capacity in key regions, especially the US, is not keeping pace, with interconnection queues showing over 2,300 GW of projects waiting to connect. The US faces a shortfall of around 9.3 GW in 2026, expected to grow to 45 GW by 2028, according to Goldman Sachs and Morgan Stanley estimates.
Meanwhile, China has significantly expanded its power generation, adding nearly 543 GW in 2025 alone, and is positioned to continue outpacing the US in capacity growth. China already produces more than twice the electricity of the US, with lower power costs for data centers and faster deployment timelines. This disparity underscores a geopolitical dimension in the AI power race, with the US and China each leading in different areas—compute and power, respectively.
For three years AI was a chip story. It quietly stopped being the binding constraint — the way it always does in a physical build-out, from the clever thing to the boring thing underneath.
When someone says AI is “only 3% of electricity,” they’re quoting consumption to make it sound modest. Capacity is where the bottleneck bites.
Implications of Power Capacity Limitations on AI Development
The constraints in power infrastructure threaten to slow or limit the scaling of AI models and data-center deployment, especially in the US. Despite significant investments—over $650 billion committed by major tech firms—physical infrastructure build-out remains a bottleneck. This could impact the timeline for AI advancements and competitiveness, as the ability to supply reliable, high-capacity power is essential for large-scale AI operations.
The geopolitical implications are notable, with the US and China each holding advantages and facing unique challenges. The US’s limited grid capacity and aging infrastructure contrast sharply with China’s rapid capacity expansion, which could influence the future leadership in AI technology and economic influence.
GOLDENMATE 1000VA/800W Lithium UPS Battery Backup & Surge Protector, White
- Long-lasting LiFePO4 Battery: Over 10 years lifespan, 5000+ cycles
- Multiple Outlets & Cooling: 8 outlets with surge protection and cooling fan
- Reliable Power Protection: Monitors voltage, prevents surges and outages
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Recent Trends in Power and Data-Center Expansion
Over the past few years, the focus in AI infrastructure has shifted from chip availability to energy supply constraints. The US has added roughly 55 GW of new power capacity in 2025, while China has added about 543 GW—almost ten times more. The US’s aging grid, with over half of coal plants built before 1980, is ill-equipped to support the surge in data-center capacity, which is experiencing a 159% increase in projects in the US alone.
Despite the large investments and ambitions, the physical build-out of power infrastructure faces long lead times. The US interconnection queue has a five-year wait time, and grid operators are advising data-center developers to reduce peak load or delay connections. These issues highlight the mismatch between financial capability and physical infrastructure readiness, a challenge that has been growing over recent years.
"Electrons are the new oil. The US must build 100 GW of new capacity annually to keep pace with China’s rapid expansion."
— Thorsten Meyer
Uncertainties in Infrastructure Expansion and Geopolitical Dynamics
It remains unclear how quickly the US can accelerate its power capacity build-out to meet AI demands, given permitting and supply chain challenges. Additionally, the extent to which China’s rapid capacity expansion will translate into sustained AI leadership is uncertain, especially considering US export controls on advanced chips. The future balance of AI power between these nations depends on resolving both energy and chip manufacturing constraints.
Next Steps for Infrastructure and Policy Responses
Key developments to watch include the US government’s efforts to fast-track power infrastructure projects, possibly through new legislation or incentives. Simultaneously, China’s continued capacity expansion and technological advancements will influence the global AI race. Monitoring grid upgrades, permit approvals, and international supply chain developments will be essential in assessing how these constraints evolve.
Key Questions
Why is power capacity more of a bottleneck than chip availability for AI?
While chip supply has been a concern, the ability to supply enough electricity at peak times is now the limiting factor for expanding data-center infrastructure. Without sufficient power capacity, new AI facilities cannot be built or operated at scale.
How does US power grid capacity compare to China’s?
China has added nearly 543 GW of power capacity in 2025 alone, significantly outpacing the US, which added about 55 GW. China’s larger, newer grid infrastructure allows faster deployment of data centers and AI infrastructure.
What are the main physical challenges to expanding power infrastructure in the US?
Permitting delays, aging infrastructure, limited manufacturing of transformers, and long interconnection queue wait times are key challenges. These physical and bureaucratic hurdles slow down the build-out of new capacity.
Could this energy constraint slow down AI progress globally?
Yes, especially in regions heavily dependent on existing, aging grids. Limited power capacity could delay large-scale AI deployment and innovation, particularly in the US and other developed countries.
What can policymakers do to address these energy constraints?
Policymakers can prioritize grid upgrades, streamline permitting processes, and incentivize new power generation projects to expand capacity quickly. International cooperation and investment in renewable energy are also potential strategies.
Source: ThorstenMeyerAI.com
