Morgan Stanley: AI Computing Demand Overwhelms Supply Long-Term, Pullbacks Are Buying Opportunities
Taylor Wilson
Morgan Stanley reaffirms its bullish stance on AI infrastructure, arguing the recent 7-percentage-point underperformance is technically driven, not fundamental — long-term compute demand still outstrips supply, making the dip a buying window.
What is this AI-stock pullback actually about?
Morgan Stanley's preferred AI-infrastructure names lagged global equities by about 7 percentage points over the past month.
This means → the sell-off was driven by technical factors — position unwinding and profit-taking — not deteriorating earnings.
In plain terms = the market is squeezing out froth, but the companies' order books and revenues haven't worsened.
Why does Morgan Stanley say "compute demand will outstrip supply"?
The bank cites three growth engines: rising AI capability + accelerating enterprise adoption + expanding capex.
Markets worry firms will cap employees' AI-token usage to cut costs. Morgan Stanley rejects this — current AI spending is only a small fraction of the productivity gains it delivers, so the incentive to cut is weak.
This reflects a view that AI spending is still "spend a little, gain a lot" for enterprises — not a burden to be slashed.
Will Chinese AI models steal global compute demand?
Morgan Stanley takes a positive view of Chinese large-language-model development.
This means → more efficient models won't shrink compute demand. Instead, they will push AI into more use cases, expanding the total global compute pie.
In plain terms = efficiency rises → more people can afford to use it → total demand goes up. The same logic as fuel-efficient cars leading people to drive more miles.
Can power gaps and construction pushback block the trend?
Morgan Stanley acknowledges three real obstacles: labour shortages, insufficient power supply, and local opposition to new data centres.
But it labels them "speed bumps," not structural barriers — they slow the pace without changing the direction.
Key figure: U.S. data centres could face a 38-gigawatt power shortfall by 2028. This means → whether new energy solutions fill that gap on schedule is the critical test of the bullish thesis.
Who does Morgan Stanley favour most?
Preferred exposures: companies benefiting from AI-infrastructure bottlenecks, semiconductor manufacturing, Chinese AI-solution providers, energy-security assets, and hyperscale cloud platforms.
Top U.S. cloud picks: Meta, Google, Microsoft, and Amazon.
This reflects a bet not just on "AI as a concept" but on the most entrenched links in the compute supply chain — cloud platforms and infrastructure.
Content is for reference only, not financial advice.