Morgan Stanley: AI Computing Power Supply Constrained, Power Gap Reaches 10% to 20%
Nashnova编辑部
Morgan Stanley strategist Michelle Weaver says AI compute is in severe undersupply, squeezed by power shortages, political pushback, and labor gaps — even after factoring in creative fixes, a 10%–20% power deficit remains, keeping compute scarce and expensive for years.
Why is compute so scarce?
Weaver told Bloomberg TV that compute is a constrained resource in severe undersupply right now.
She pinpoints three bottlenecks: power shortages, political resistance, and labor gaps — all three hit at once and won't ease for years.
This means → compute isn't just a "build it and it comes" problem. The supply side is locked; prices and lead times stay elevated.
How big is the power gap?
Even after counting Bitcoin-mine conversions, fuel cells, and other unconventional power fixes, Weaver estimates a 10%–20% power shortfall remains.
In plain terms = every creative workaround is already in the math, and supply still falls one-to-two-tenths short — this gap doesn't close quickly.
Data-center construction also faces a labor shortage, further slowing the buildout.
What's the good news on demand?
Among S&P 500 companies, 25% can now quantify measurable returns on AI spending — up from 14% a year ago. Enterprises are starting to make real money from AI.
On the funding side, Nvidia and Wall Street firms have arranged $500 billion in AI-infrastructure financing — capital is not the constraint.
This means → AI has no shortage of demand or money. The real chokepoint is supply — power and people.
How could politics make things worse?
As U.S. midterm elections approach, anti-data-center sentiment is rising: voters worry about higher electricity bills and impacts on water and air quality.
Weaver says the industry has mitigation plans, but the issue could intensify during the campaign sprint, adding uncertainty to permits and site selection.
This reflects a broader shift — compute expansion is no longer just a tech-and-capital problem. Public sentiment and ballots are becoming real constraints.
What does this mean for investors?
Constrained supply + accelerating demand = compute stays scarce and high-value for years.
This means → companies that control compute resources — data centers, power providers, GPU makers — sit in a seller's market with pricing power.
But the pace of payoff hinges on two gates: power and politics — the faster supply can expand, the faster returns arrive.
Content is for reference only, not financial advice.