Analysts Revised US Data Center Power Shortfall Outlook
New projections suggest a massive gap between energy capacity and demand through 2029.
Updated on Sept. 21, 2026 in Data Centers

Live Poll
Is the rapid expansion of AI data centers creating an unsustainable demand on the power grid?
Morgan Stanley analysts have raised estimates for United States power shortfalls, warning of a potential 38-gigawatt deficit by 2028. This comes as rising demand from advanced hardware architectures drives unprecedented consumption rates.
Why it matters
Hardware efficiency gains are inadvertently fueling higher total energy usage due to Jevons paradox, where cheaper operational costs drive increased consumption density. As data centers pack more power into existing footprints, the grid faces a capacity gap of 30 to 40 per cent.
Nvidia's Vera Rubin NVL72 server rack integrates 72 GPUs, with racks producing heat equivalent to 10 to 25 ovens. Meanwhile, future architectures like Rubin Ultra in 2027 and Fenyman in 2028 are expected to further intensify rack power density.
The players
Morgan Stanley
This is a global financial services firm that provides investment banking, securities, and wealth management services to institutional and individual clients.
Nvidia
The company is a major technology manufacturer known for designing graphics processing units and hardware architectures for artificial intelligence and data centers.
Elon Musk
He is a prominent business executive and entrepreneur who serves as the leader of several major technology and energy-related corporations.
APR Energy
This organization specializes in providing rapidly deployable, large-scale power generation solutions to utility and industrial customers globally.
The details
System designers are utilizing hardware efficiency gains to increase computing capacity rather than reducing total electricity needs. As a result, the industry expects total IT power demand to increase fivefold between 2025 and 2028 as processors become more powerful.
Timeline
Q3 2022: Nvidia released the Hopper chip family.
July 2026: Elon Musk acquired APR Energy for $1 billion.
2025-2028: IT power demand is expected to increase fivefold.
H2 2027: Rubin Ultra chip architecture is scheduled for launch.
2028: Target year for current US power shortfall estimates.
The Tech Race
This power crisis follows the pattern set by Jevons paradox, where increases in hardware efficiency lead to greater total resource consumption. As companies chase higher compute density, the industry is increasingly forced to seek independent energy solutions.
The projected 107-gigawatt shortfall by 2029 could lead to higher utility costs for residential consumers as data centers compete for grid capacity. Users may also experience slower adoption of high-performance digital services if power constraints limit the rollout of new processing hardware.
The takeaway
The rapid expansion of AI hardware is fundamentally outstripping existing grid capacity, forcing tech giants to move into the energy generation business. Consumers should expect energy costs to remain volatile as major industrial players prioritize power stability to maintain their computing lead.
What happens next
The Rubin Ultra chip architecture is scheduled for release in the second half of 2027, followed by the Fenyman chip architecture in the second half of 2028.
Further reading
Learn more about the infrastructure challenges facing the industry in our Data Centers section.
Source note: This article includes information reported by Financial Times News.
Live Poll
Is the rapid expansion of AI data centers creating an unsustainable demand on the power grid?










