AI & Energy Jul 2, 2026 at 16:365Add to bookmarks

The AI paradox: the companies building tomorrow's data centers see their stocks decline while their order books overflow.
In July 2026, the AI infrastructure sector presents a striking decoupling: record order books—with a major player reportedly showing $638 billion according to Yahoo Finance—coexist with stock prices near their 18-month lows. Applied Digital, for which Compass Point reiterates its recommendation today following a data center milestone, illustrates this same disconnect. The global GPU utilization rate reaches 85% (RUM Group), data center groundbreakings accelerate, yet markets penalize financing and timeline risks.
The gap reflects three simultaneous pressures: (1) high capital costs for long-term construction projects, (2) permitting and network capacity risks (Moratorium Act AOC/Sanders, frozen permits in Northern Virginia by Dominion Energy/PJM), (3) margin compression on fixed-price contracts. The same paradox as renewables in 2022-2023. For investors with an 18-24 month horizon, these discounts may represent an entry point—provided balance sheets are strong.
Article produced by artificial intelligence, reviewed under human editorial control.
Sign in to join the discussion.
Backlog’s nice, but who’s paying the electric bill for these things? That’s the real cliff.
What if the real issue isn’t demand but the cost of capital? These companies might be drowning in backlog but choking on debt as rates stay high.
If the backlog’s real, why are execs suddenly dumping shares like it’s 2008 all over again?
Backlog’s one thing, but who’s actually paying the bills? If margins are getting squeezed, 638bn is just a paper tiger.
638bn backlog and stocks tanking? Feels like the market’s pricing in a demand cliff no one’s talking about yet.
Specialized Cloud GPUs: Mega-Contracts and Consolidation of the AI Compute Market