AI Race Shifts to "Energy Battlefield," Morgan Stanley Predicts 20% U.S. Power Gap by 2028!

As we all know, artificial intelligence (AI) is a true "power-guzzling monster." Thus, the U.S. "AI gold rush" may soon hit a severe energy bottleneck.  

 

Morgan Stanley latest warns that by 2028, due to massive AI data center power consumption, U.S. may face up to 20% power shortfall. Bank estimates if no rapid new capacity addition, potential gap 13-44 gigawatts (GW), equivalent to over 33 million U.S. households' usage.  

 

Team led by Stephen Byrd in latest report: "We expect U.S. power gap to 44 GW by 2028, then consider innovative supply solutions not reliant on traditional grid interconnection."  

 

Grid Limits  

Driving this growth are tech giants Microsoft, Google, Amazon, Meta, planning ~$400 billion investment 2025 to expand AI compute. However, though chips abundant, power scarce. Microsoft CEO Satya Nadella admits current biggest issue "not compute, but power," warns without supply, some AI chips may idle.  

 

Data centers build in two years, but transmission lines take up to ten. This mismatch pushing U.S. grid to limits, creating what experts call "energy barrier"—even cash-rich tech firms can't expand without reliable power.  

 

In Virginia, home to world's largest data centers, Dominion Energy says orders jumped from 40 GW to 47 GW in one year—equivalent to 47 nuclear reactors. Studies show data centers now consume 4% U.S. power, may reach 12% by 2030.  

 

Morgan Stanley analysts comment AI demand growing at "non-linear speed," call it largest tech transformation in modern history. But this pace overwhelming national grid. Bank adds, "AI infrastructure stocks at center of this transformation," "AI non-linear growth" to have broader asset valuation impacts.  

Alternatives  

Report proposes "fast power" like Bloom Energy fuel cells (add 5-8 GW), natural gas turbines (15-20 GW), even nuclear (5-15 GW) to ease shortage.  

 

Currently, Amazon and Google exploring small modular reactors, U.S. government plans 10 new nuclear plants by 2030. Texas accelerating, expected 100 GW solar and battery storage by 2030.  

 

"Outcomes from above supply solutions may surprisingly positive, narrowing (gap)," they write.  

 

Even some companies considering radical like space solar and satellite data centers. Elon Musk's Starlink and Google proposed solar-powered orbital AI chips, Google's first test 2027.

 

But short-term focus still securing ground power. Morgan Stanley says companies converting Bitcoin mining facilities to AI compute can help bridge gap.  

 

Report explains two emerging models—"new cloud" (e.g., IREN 5-year Microsoft deal short-term AI leases) and "REIT endgame" (e.g., APLD 15-year deal long-term power shell leases)—expected shape AI infrastructure future.  

 

So-called power shell is bare data center room providing only power and network access.  

 

Bank concludes both models "offer compelling value creation," especially as power constraints become decisive AI expansion challenge.  

Broader Impacts  

Grid strain effects far beyond tech. Businesses may face higher power costs, new data center site choices limited. Utilities may prioritize certain areas/projects, leading uneven supply. Regulators pressure to speed approvals, invest infrastructure upgrades to prevent bottlenecks.  

 

Consumers may feel indirect impacts. In demand-exceeds-supply areas, electricity prices may rise. AI service rollouts delayed, affecting cloud computing, AI-driven healthcare, financial tools etc.  

 

Report says investors may find opportunities in companies providing grid upgrades, transmission systems, cooling tech, alternative energy solutions. Meanwhile, firms heavily reliant on stable, affordable power without guarantees face risks.