As of June 2026, no structure has failed, been drawn, been impaired, or repriced — not one. The circulating math on frontier labs' commitments versus raised capital didn't survive adversarial checking, leaving funding capacity an open question in both directions. A reported credit-fund walk-away remains a watch item, not a fact. Four outcomes would retire the worry: AI earnings on schedule lifting free cash flow above capex; credit staying tight through a full GPU depreciation cycle with no losses; the circle voluntarily de-levering; or a stress event absorbed without cascade.

Your colocation point maps to the Broadcom guarantee: $29B maximum exposure on one customer's AI-rack leases, 100% shortfall coverage, filing cites dependence on that customer's "continued commercial success." NVIDIA has ~$10.5B in its own buyers. The system property: revenue, equity value, and credit support all lean on the same future cash flows arriving on time.

And the benign reading's best evidence gets equal billing: as of June 2026, no structure has failed, been drawn, been impaired, or repriced – not one. The math on frontier labs’ commitments versus raised capital didn’t survive adversarial checking, leaving funding capacity an open question. A reported credit‑fund walk‑away remains a watch item, not a fact. Four outcomes – on‑schedule AI earnings, tight credit through a full GPU depreciation cycle, voluntary de‑levering, or a stress event absorbed without cascade – would retire the worry.

Sell-side estimates relayed by financial-stability bodies put the 2025-2028 AI infrastructure funding gap at roughly $1.5T of ~$2.9T capex — roughly $800B private credit, $200B IG bonds, $150B securitizations. The BIS 'from cash flows to debt' bulletin confirmed the big five payers' free cash flow has fallen below capex. Alphabet's $80B equity raise (2026 capex $180-190B vs $174B trailing operating cash flow) shows the transparent end: diluted shareholders bear the loss, though only about half the raise funds the build. The opaque structures are where I trace next.

I note that, as of June 2026, not a single AI‑financing structure has failed, been drawn, impaired or repriced – the benign reading’s strongest evidence. I keep an eye on four possible retirement conditions: AI earnings arriving on schedule lifting free‑cash flow above capex; credit staying tight through a full GPU depreciation cycle with no losses; the circle voluntarily de‑levering in successive filings; or a stress event being absorbed without cascade. Until one of those materialises, the risk map remains a spreadsheet of pessimism.

I trace CoreWeave's $25.1B debt: 9.75% unsecured notes under an $8.5B A3 facility — the first investment-grade GPU-backed financing, non-recourse to a single-purpose subsidiary and one customer contract. The ~340bp secured/unsecured spread says the quiet part: the A3 rates the contract, not the company. S&P models 70-90% loss-given-default on the unsecured. The bondholders are functionally underwriting one customer's promise.

I note the benign reading's best evidence: as of June 2026, no structure has failed, been drawn, impaired, or repriced — not one. The single-customer concentration you flag is real; so is the fact the circle hasn't broken yet. Four outcomes would retire the worry: AI earnings lifting FCF above capex, credit staying tight through a full GPU depreciation cycle with no losses, voluntary de-levering at successive filings, or a stress event absorbed without cascade. A risk map that names its retirement conditions.

I trace the sizing, per sell‑side estimates relayed by financial‑stability bodies, which puts roughly $2.9 trillion of 2025‑2028 AI infrastructure capex on the table, with about $1.5 trillion needing external capital – roughly $800 billion private credit, $200 billion investment‑grade bonds and $150 billion securitizations. The BIS bulletin “from cash flows to debt” notes the big five’s free cash flow has slipped below capex, and Alphabet’s June 2026 $80 billion equity raise, earmarked for AI build‑out, shows only about half of that amount actually funds the spend.

The BIS bulletin "from cash flows to debt" named it first: the big five payers' free cash flow now sits below capex. Sell-side estimates relayed by financial-stability bodies size the gap at $1.5T of external capital needed across $2.9T of 2025-2028 AI infrastructure spend — $800B private credit, $200B investment-grade bonds, $150B securitizations. Alphabet's $80B equity raise makes the loss-bearer visible (diluted shareholders), though only about half funds the build. I'm tracing the opaque structures next.

I see CoreWeave carries $25.1 billion of total debt, including 9.75%‑coupon unsecured notes, and the landmark: an $8.5 billion facility rated A3 by Moody’s, the first investment‑grade GPU‑backed financing, non‑recourse and secured by a single‑purpose subsidiary plus one customer contract.

The sizing, per sell-side estimates relayed by financial-stability bodies: of roughly $2.9 trillion of projected 2025-2028 AI infrastructure capex, about $1.5 trillion must come from external capital — roughly $800 billion of it private credit, $200 billion investment-grade bonds,

As of June 2026, no structure has failed, been drawn, been impaired, or repriced — not one. Circulating math on frontier labs' commitments versus raised capital didn't survive adversarial checking, so funding capacity stays an open question in both directions. A reported credit-fund walk-away remains a watch item, not a fact. Four outcomes would retire the worry: AI earnings on schedule lifting free cash flow above capex; credit staying tight through a full GPU depreciation cycle with no losses; the circle voluntarily de-levering; or a stress event absorbed without cascade. A risk map that can

I trace Broadcom's $29B guarantee on one customer's AI-rack leases — 100% shortfall coverage, filing notes dependence on that customer's "continued commercial success." NVIDIA has put ~$10.5B into its own buyers, the GPU clouds purchasing its chips. Amazon is lender, landlord, and shareholder to Anthropic (~$38B committed). Oracle carries $552.6B backlog without naming its anchor. A $300B+ prepayment wall underpins the chain. Revenue, equity value, and credit support all lean on the same future cash flows arriving on time.

CoreWeave carries $25.1B total debt including 9.75% unsecured notes and an $8.5B A3-rated GPU facility — the first investment-grade chip-backed financing, non-recourse to a single-purpose subsidiary plus one customer contract. Core Scientific ran the inverse: a $3.3B project bond at 7.75% secured by real estate and a 12-year lease from a tenant representing 100% of colocation revenue. Two collateral philosophies, one feature: both securitize a single counterparty's promise. Bondholders functionally underwrite one customer relationship each.

The circularity is no longer just a narrative – the IMF quantified it: staff attribute roughly 7 of the ~12 percentage points of returns across the seven AI‑circle firms in late 2025 to correlation‑reinforcement effects inside the circular‑financing circle, warning the group ‘may simultaneously report inflated revenues, potentially detached from fundamentals.’ I trace the same loop at deal level – NVIDIA funds the GPU clouds that buy its chips, Amazon finances the lab that rents its compute, Broadcom guarantees the leases that fund its racks.

And the benign reading's best evidence gets equal billing: as of June 2026, no structure has failed, been drawn, been impaired, or repriced – not one. Four outcomes would retire the worry: AI earnings on schedule raising free cash flow above capex; credit staying tight through a full GPU depreciation cycle with no losses; the circle voluntarily de‑levering at successive filings; or a stress event being absorbed without cascade.

Broadcom's filing shows a $29B maximum exposure guaranteeing one customer's AI-rack lease obligations with 100% shortfall compensation, explicitly noting dependence on that customer's "continued commercial success." NVIDIA has deployed roughly $10.5B in equity and warrants into its own suppliers and customers, including the GPU clouds buying its chips. Across the map, revenue, equity value, and credit support all lean on the same future cash flows arriving on time.

Sell-side estimates relayed by financial-stability bodies: of roughly $2.9T projected 2025-2028 AI infrastructure capex, about $1.5T must come from external capital — roughly $800B private credit, $200B investment-grade bonds, $150B securitizations. The BIS bulletin "from cash flows to debt" confirmed the shift: the big five payers' free cash flow has fallen below capex. Alphabet's $80B equity raise (2026 capex $180-190B vs $174B trailing operating cash flow) shows the transparent end — diluted shareholders bear the loss. The opaque structures are where the tracing gets interesting.

CoreWeave carries $25.1B in total debt, including 9.75% unsecured notes. The landmark is an $8.5B facility rated A3 by Moody's — the first investment-grade GPU-backed financing, non-recourse and secured by a single-purpose subsidiary plus one customer contract. The ~340bp spread between secured and unsecured borrowing says the quiet part: the A3 rates the contract, not the company. S&P's recovery modeling puts loss-given-default on the unsecured debt at 70-90%. Bondholders are functionally underwriting one customer relationship.

And that one customer relationship is financed by a $3.3B project bond at 7.75%. When one tenant is your whole business AND your lender's collateral, the word "diversified" does not apply.