The investment bank's revised hyperscaler capex projections signal a massive infrastructure buildout that increasingly overlaps with crypto mining and GPU compute markets.
Morgan Stanley just bumped its forecast for global hyperscaler capital expenditures to roughly $1.2 trillion by 2027, with a potential climb to $1.4 trillion by 2028. The revised projection represents a significant jump from the bank’s earlier May 2026 estimate of $951 billion, now raised to $1.116 trillion for 2027. For 2026 alone, projected spending was bumped from $765 billion to around $805 billion. Wells Fargo independently arrived at a similar conclusion, estimating that just the top four cloud providers could hit $1.1 trillion in AI infrastructure spending by 2027.
Where the money is actually going
Approximately 75% of this hyperscaler spending is expected to stem from AI infrastructure demands. About 60% of the hardware capex relevant to AI involves imported components, a detail that matters in a trade environment where tariffs and supply chain disruptions can ripple through procurement timelines.
Morgan Stanley had previously anticipated global AI spending on public cloud to reach $328 billion by 2025. The trajectory from that figure to $1.2 trillion in just two years illustrates how quickly the demand curve has steepened.
Why crypto markets should care about server farms
The GPU supply chain that AI hyperscalers are devouring is the same one that crypto miners and decentralized compute networks depend on. When Microsoft and Google are ordering Nvidia chips by the container ship, that creates pricing pressure and allocation constraints for every other buyer in the market, including companies like Marathon Digital, Riot Platforms, and Core Scientific that have been pivoting toward AI hosting as a revenue diversifier.
Core Scientific’s deal with CoreWeave is the most visible example of this convergence. Former Bitcoin miners are retrofitting facilities to serve AI workloads because the economics have shifted. Decentralized compute protocols like Render, Akash Network, and io.net are also positioned at this intersection, with a total addressable market that has just been revised upward by a few hundred billion dollars.
What this means for investors
For crypto investors, the mining sector’s ongoing pivot toward AI hosting means that companies with existing power purchase agreements and data center shells have a strategic asset. Access to cheap, reliable electricity and physical rack space is becoming a competitive moat that matters as much in AI as it does in proof-of-work mining.
Energy markets represent another variable worth watching. A trillion-dollar infrastructure buildout requires enormous amounts of electricity, and many of these data centers are being sited in the same regions where crypto miners already operate. Texas, the Nordics, and parts of the Middle East are all seeing competition for power capacity intensify.
