US AI Data Centres Face 'Six NYC' Electricity Shortfall: UK Grid Implications?
Executive summary
A new report by Morgan Stanley suggests US data centres, particularly those supporting AI, face a significant electricity shortfall, potentially reaching 107 GW by 2029. This is attributed to the rapidly increasing power demands of advanced AI chips, which are outstripping power generation capacity.
Reporting based on ft.com
Why it matters
While directly focused on the US, this article highlights the immense and rapidly escalating power demands of AI infrastructure. For the UK, this serves as a critical indicator of future energy needs, potentially informing strategic planning for grid capacity, generation investment, and the siting of new data centres to avoid similar shortfalls and ensure energy security amidst AI adoption.
Sector impact
Analysis by AI Energy Intelligence UK
The article reports that advanced AI chips, such as Nvidia's Vera Rubin and Rubin Ultra architectures, threaten to increase IT power demand by fivefold from 2025 to 2028. Morgan Stanley estimates a potential cumulative US power shortfall of 107 GW by 2029, even with some 'time-to-power' solutions, driven by the Jevons paradox where increased chip efficiency leads to greater overall power consumption.
The projected power shortfall in the US, described as equivalent to multiple New York City's baseload demand, illustrates the potential for AI infrastructure to strain national grids. This raises concerns for the UK regarding its own energy security, as a failure to match AI growth with adequate power generation could lead to grid instability or reliance on less secure energy sources.
The article suggests that if the power shortfall cannot be closed, the expected pace of AI improvement and adoption may not fully materialise, impacting the profitability and growth trajectories of tech companies. It also identifies investment opportunities in companies that can provide viable energy solutions, including 'powered shell providers', novel generation companies, and equipment suppliers.
Key statistics
Figures as reported by ft.com. See original source for context.
Quotations
"There’s quite a crisis of power. [. . .] The consensus at this point is that there will be a significant power shortfall next year. So, not like distant future. I believe the consensus estimate among analysts who follow the AI space very closely is that there will be at least a 15-gigawatt shortfall of power in 2027 for AI chips. The rate at which AI chips [are] being produced has been rising incredibly rapidly. They’re rising on the order of 40 to 50 per cent a year. But power available outside of China has been rising at 10 per cent to 20 per cent a year. Obviously, the faster-rising thing will eventually overwhelm the slower-rising thing."
"New York City demands 5.5-6 GW of baseload power: we are short power by six New York Cities."
"if we cannot close the power shortfall, the expected pace of improvement and adoption may not fully materialise in line with market expectations."
Long-term implications
The long-term implications point to a persistent and growing challenge for national energy systems to keep pace with AI's power requirements. This suggests a future where grid capacity and reliable energy generation, including innovative 'time-to-power' solutions like fuel cells and behind-the-meter nuclear plants, become critical limiting factors for technological advancement and economic growth, potentially necessitating significant infrastructure investment and strategic policy interventions.
Frequently asked questions
What is the main finding of the article?
The article reports that US data centres face a substantial electricity shortfall, estimated to reach 107 GW by 2029, primarily due to the rapidly increasing power demands of advanced AI chips.
How much electricity is a 'New York City' in this context?
The article states that New York City demands 5.5-6 GW of baseload power.
What is Jevons paradox and how does it relate?
Jevons paradox is the phenomenon where improved efficiency makes a resource cheaper to use, increasing consumption and undoing initial savings. In this context, more efficient AI chips lead to data centres packing more processing power, thus increasing overall electricity consumption.
Explore related tools
Original source
This story summarises reporting from ft.com. Read the original for full context.
Read on ft.comRecommended reports
Independent UK research that expands on the themes in this story.
The environmental cost of AI-generated answers versus conventional search, benchmarked for UK decision-makers. An independent, evidence-based comparison of electricity, water and carbon per query across Google, Gemini, ChatGPT, Copilot, Perplexity and Claude — with forecasts to 2035.
Read the reportThe definitive UK view of AI's electricity, grid and infrastructure impact in 2026. An independent, evidence-based report on demand growth, data centre build-out, AI Growth Zones and the trajectory to 2035 — for UK policymakers, operators and investors.
Read the report
A structured briefing on how artificial intelligence infrastructure is reshaping Britain's power network — from substation bottlenecks to regional investment opportunities and long-term grid risk.
Read the reportGet the UK AI Energy briefing
Analysis on AI, electricity and the UK grid, straight to your inbox.
Evidence behind this story
Our reporting sits on top of the UK AI Energy Index and Data Centre Tracker — a sourced, dated record of AI and data-centre electricity demand, data-centre development and grid pressure.