Morgan Stanley Pushes Some U.S. Data Center Capacity to 2029
The bank says election-related changes are delaying about 5 GW of planned U.S. capacity, though its AI spending outlook remains intact.
Updated on Nov. 11, 2026 in Data Centers
Live Poll
Do you expect AI to improve your daily life over the next few years?
Morgan Stanley expects operators to bring about 5 GW of planned U.S. data center IT capacity online in 2029 instead of 2028. The delayed capacity is concentrated in Texas and Ohio and represents about 4% of planned U.S. capacity.
Why it matters
The bank characterizes the change as a postponement rather than a cancellation, with projects already under construction not interrupted. Its view is that election-related changes are affecting construction timing rather than its assessment of AI demand.
Morgan Stanley's revised forecast puts 2028 U.S. data center additions at 19.9 GW, down from 24.9 GW, while 2029 additions rise to 27.6 GW from 22.7 GW. Globally, the bank lowered its 2028 capacity forecast to 219 GW from 224 GW, and it is not yet clear whether the revised timing will hold.
The players
Morgan Stanley
A global investment bank and financial services firm whose research division publishes equity and sector forecasts closely watched by investors tracking the semiconductor and data center markets.
The details
Morgan Stanley attributes the delays to election-related changes and estimates that about 20% of U.S. projects will slip by a year, compared with about 10% of global projects. Its total AI capital-spending forecasts remain intact.
Timeline
Nov. 6, 2026: Median semiconductor stock covered by Morgan Stanley is 53% below its 52-week high.
Nov. 8, 2026: Morgan Stanley publishes its outlook on the delayed capacity.
2028: Original expected year for the 5 GW of capacity to come online.
2029: Revised expected year for the 5 GW of capacity to come online.
Market Landscape
The revision shifts timing within Morgan Stanley's own forecast, moving about 5 GW from 2028 into 2029 while its 2027 projection stays unchanged. The bank's analysis adds to a broader sell-off in chip stocks, with the Philadelphia Semiconductor Index down about 40% since mid-June.
Delays in data center construction can affect the pace at which new cloud and AI computing capacity becomes available to businesses that rely on it. Projects already under construction are expected to continue, so near-term service availability is less likely to be affected than longer-range expansion plans.
The takeaway
Forecast timing matters as much as totals, so readers tracking AI infrastructure should watch whether capacity slips are isolated or spread across more projects. The shift suggests that headline capacity figures can move year to year even when long-term demand expectations hold.
Further reading
For broader context on infrastructure buildouts, see Data Centers coverage.
Live Poll
Do you expect AI to improve your daily life over the next few years?
