RegLayer: 47+ large US data center projects were delayed, withdrawn, or blocked in 2026; Bernstein says ~35% of the planned US data center pipeline is credible
A vast bet on the future of the technology is running into the unforgiving economics of the physical world
Financial Times
Context & Ripple Effects
The evidence has accumulated through 2026: SynMax’s early estimate of widespread 2026 delivery delays was followed by JPMorgan’s finding that much of 2027’s planned capacity had not entered construction. September coverage then tied a large share of delayed or blocked projects to local opposition.
RegLayer’s tally and Bernstein’s credibility screen turn those scattered schedule and permitting problems into a pipeline-quality question. The issue is no longer only whether developers announce capacity, but whether sites can clear local constraints and progress into buildable supply.
First-order effects
Bernstein’s assessment narrows the set of US data-center proposals that developers, lenders and capacity buyers can treat as dependable future supply.
Cloud and AI customers with capacity tied to delayed developments must place greater value on projects already under construction or with fewer permitting obstacles, tightening competition for credible supply.
Data-center financing is pushed toward demonstrable site readiness and construction progress rather than announced pipeline scale, as the gap between planned and credible capacity becomes explicit.
Third-order effects
If these execution constraints persist, US AI-compute expansion will be shaped less by headline capital commitments than by control of developable, approvable sites and the ability to deliver them.
The build-out is becoming a test of whether compute can function as financeable infrastructure when local permitting, construction timing and pipeline credibility are inseparable from demand forecasts.
The trend:AI infrastructure is moving from an announced-capacity race toward a site-readiness and execution-risk market.
Cracks are appearing in America's debt-laden $10tn AI infrastructure boom as project delays, public backlash and shaky investor confidence bring a reckoning for the data centre build-out underpinning the US economy https://www.ft.com/...
Investors are DUMPING AI infrastructure debt: Bonds financing major US data centers are sliding below par, with Hyperion debt falling to just under 92 cents on the dollar, Hut 8-backed projects trading at 94 and 92 cents, and Blackstone-owned Project Excalibur sinking below 88 ce…
Why an AI bust could hurt far more than the dot-com crash did. In the words of the man known as the “Dean of Valuation”, Professor Aswath Damodaran, of NYU Stern School of Business,
Investor fears over the cost of the AI infrastructure buildout are SKYROCKETING: CoreWeave, $CRWV, dropped -20% over the first 3 sessions of the week, and then rebounded +5.9% on Thursday after Nvidia, $NVDA, CEO Jensen Huang expressed confidence in the company. However, $CRWV cr…
Did cap structure arb 2007 to 2016. Converts .... On and off delta based on views L/s credit within same cap structure Distressed Long bonds vs short puts Short bonds vs long calls Done it all Crwv probably one of greatest cap structure arb trade setups ever. Do not think it fail…
'The broader question is whether companies facing debt trouble like CoreWeave, Oracle & SoftBank are the industry's weak links or are harbingers of deeper problems in the A.I. build-out.' '"The market is on a knife's edge..."' https://www.nytimes.com/...
Weak Links Start to Show in A.I. Borrowing https://www.nytimes.com/... via @NYTimes “The market's waking up to the fact that there is potentially a mismatch between the money they have to commit today and borrow today and the ability to generate revenues down the road,” he added.
15% is getting well into junk bond territory (I don't know what their actual ratings are, but the 2007-8 financial crash had all three rating agencies calling things AAA when they were totally toxic assets.)