The AI cracks are starting to become very prevalent, and might be...
KKGB@INArteCarloDoss
87 views
Sep 30, 2026
~2 min read
Advertisement
1
The AI cracks are starting to become very prevalent, and might be the reason both the industry stalwarts have started to flout the need to be brought under the state umbrella (bailouts?) and grifters touting nonsensical shit like infinite ROI.
1/10
1/10
2
We all understand the dual issue of LLM uncertain profitability and hyperscalers’ negative net cash flows but what is bringing down the house of cards is credit. And I really am not referring to Oracle CDS, it’s a shit B/S and market has already given it too long a rope
2/10
2/10
3
The mega data centres projects are hitting a permitting and interconnection wall. New Mexico State Land Office repeatedly denied permits for the Energy Transfer natural gas pipeline and Texas temporarily halted major data center grid approvals pending ERCOT energy audits
3/10
3/10
4
They are also hitting industrial supply chain bottlenecks. High-volt transformers now face 3 to 4 y delivery backlogs and specialized switchgea past 80 w. Finally local opposition (water rights, noise, electricity rate spikes) and regulatory scrutiny are surging
4/10
4/10
5
All of this is creating debt structure frictions. SPVs are usually a mix of private credit and syndicated loan books (some still sitting partly on underwriters B/S). Delays are going to raise legal hell with regards to tenant obligations and their B/S capacity to fulfil them
5/10
5/10
6
100s of billions sit on private credit and bank syndicates B/S (assuming hyperscalers guarantee). If tenant legal disputes multiply, rating downgrades hit the SPV debt, forcing lenders to mark down assets and tighten underwriting terms xss the entire digital infrastructure
6/10
6/10
7
If negative free cash flow persists alongside delayed operational capability, tech companies will be forced to renegotiate, pause, or write off large portions of their build-outs—triggering an industry-wide consolidation in AI infrastructure.
7/10
7/10
8
Over the past two years, debt-financed Capex shifted from <10% to over 33%.
For a combined debt load of circa $ 800 b a market-wide repricing phase of 100–150 bps across corporate debt curves would add $ 8-12 b in net cash interest expenses across the sector
8/10
For a combined debt load of circa $ 800 b a market-wide repricing phase of 100–150 bps across corporate debt curves would add $ 8-12 b in net cash interest expenses across the sector
8/10
9
The corporate bond market is facing real structural limits in absorbing this volume of debt:
AI-related debt now accounts for roughly 30% of net new US IG bond supply, way > concentration thresholds which is driving spreads widening. And that’s before the brewing crisis
9/10
AI-related debt now accounts for roughly 30% of net new US IG bond supply, way > concentration thresholds which is driving spreads widening. And that’s before the brewing crisis
9/10
10
If off-grid power delays and grid capacity bottlenecks push out the operational start dates of multi-billion dollar campuses, revenue generation stalls while interest expenses compound. then debt will be priced as traditional, capital-intensive utility infrastructure debt
10/10
10/10