By Maksym Misichenko · ZeroHedge ·
What AI agents think about this news
The panel generally agrees that the $35B 'Project Big Sky' deal, while innovative, carries significant risks. The deal's structure, which relies on synthetic credit engineering and off-balance-sheet financing, creates potential vulnerabilities such as cliff risk, moral hazard, and systemic vulnerabilities in case of a slowdown in AI capex ROI or lease payment defaults.
Risk: Rapid GPU depreciation leading to lease cash flow evaporation and a collapse of the SPV structure, leaving credit holders with obsolescing silicon and potential mark-to-market losses.
Opportunity: None explicitly stated by the panel.
This analysis is generated by the StockScreener pipeline — four leading LLMs (Claude, GPT, Gemini, Grok) receive identical prompts with built-in anti-hallucination guards. Read methodology →
Apollo And Blackstone Raise $35 Billion For Anthropic In One Of The Biggest Ever Private Credit SPV Deals
Back in January, when we profiled Meta's landmark $27.3 billion SPV deal named "Beignet" for the Hyperion data center located in Louisiana, in which Blue Owl provided the private credit, we said to "expect hundreds of billions of these in 2026."
As a reminder, META is already neck deep in off-balance sheet debt. Here is a schematic of its $27.3 billion SPV with Blue Owl "Project Beignet" for the Hyperion data center. None of this touches META's balance sheet.
Expect hundreds of billions of these in 2026 https://t.co/794EgSiiZ9 pic.twitter.com/7hMyVW6Lno
— zerohedge (@zerohedge) January 29, 2026
Fast forward five months when we now read that Apollo and Blackstone have finalized a $35BN private credit deal that will help finance Anthropic’s growth plans, even as traditional "banks are choking" on the amount of AI debt they have to issue.
The two private credit giants - which in a parallel universe are struggling with soaring redemption requests and gating retail investors in their private credit BDCs as documented here extensively in recent months - led the financing, one of the largest private credit deals completed, which will fund Anthropic’s purchase of Alphabet-developed chips.
The deal, dubbed project “Big Sky”, comes amid concerns that the AI frenzy has overheated the broader market. Shares in chipmakers rebounded on Monday after tumbling last week, led by Broadcom’s fall in market value. It also adds to a deluge of chip-backed loans that sparked debate over how quickly graphics processing units would depreciate as AI technology evolves.
In this type of financing structure, a special-purpose vehicle raises capital through a mix of debt and equity to purchase the chips, which are then leased to a customer, in this case Anthropic. The debt is primarily backed by the resulting lease payments, along with the unknown long-term value of the chips. In this case, the $35 billion debt facility was structured across three tranches. The senior layers — the $6 billion notes dubbed A1 and $24 billion of A2 notes — are backed by Broadcom, allowing the debt to secure lower borrowing costs aligned with Broadcom’s strong credit profile. The notes received private ratings in the mid-investment grade tier.
The transaction wrapped up days after Alphabet completed one of the largest equity offerings in history, as it looks to raise $85bn to fund Google’s AI build-out, and as SpaceX prepares for a flotation that could raise a record $86bn. Anthropic also announced it had confidentially filed for an IPO shortly after its blockbuster $65bn private financing round.
As discussed previously, the AI borrowing spree has reached beyond traditional US capital markets, where AI is expected to raise $400 billion in debt, rising to over $1 trillion through 2028 to meet roughly $1.8 trillion in capex needs over the next two years, according to Morgan Stanley...
... with Amazon raising C$14bn (US$10bn) on Monday in the largest ever Canadian dollar bond sale.
Similar to Meta's Beignet deal, Anthropic’s deal with Apollo and Blackstone relies on a complex structure that private investment groups routinely use to finance start-ups with backing from blue-chip companies. A special purpose vehicle formed by Apollo’s Atlas SP Partners raised the debt and equity, with lease agreements for the chips ultimately supporting the value of the transaction.
Per the FT, Apollo and Blackstone structured the loan across three tranches, with interest payments on the two senior segments backstopped by Broadcom. The chipmaker is making the so-called tensor processing units, or TPUs, with Google. Its agreement to provide support if Anthropic misses an interest payment helped vastly reduce the costs on the debt.
The two senior portions of the debt were split between banks and investors. Some $6bn of so-called A1 notes were sold to banks with an interest rate 1% over Treasuries. A further $24bn of A2 notes were sold on to investors in asset-backed credit markets, priced with a yield of 5.75 per cent. Buyers of the A2 tranche included institutional investors like Apollo’s Athene insurance arm, which favors high-quality debt to back its long-term liabilities.
The $4.5bn of junior debt, which is not supported by Broadcom and therefore exposes lenders more acutely to Anthropic, carried an interest rate of 8.5%. Investors were also offered an original issue discount of 98 cents to 99 cents on the dollar depending on cheque sizes. In other words, without the implicit guarantee from an investment grade guarantor - like Broadcom in this case - the cost of capital is roughly double.
In addition to the debt, Apollo’s Atlas SP Partners’ structured-finance unit provided $800 million in equity, meaning it’s effectively the owner of the SPV.
A key feature of the deal is Broadcom providing a “residual value support” agreement. That means that if Anthropic fails to make the lease payments for a certain period of time, the SPV will sell the chips to pay back the debt investors. If the value of the chips doesn’t make the debt investors whole, then Broadcom will make up the shortfall for 100% of the value owed to the A1 and A2 investors.
This type of residual value feature has been used in another mega debt deal, though it financed the construction of a data center rather than chips. As noted above, Meta provided a similar protection for the value of its Hyperion facility in Louisiana - a transaction that Morgan Stanley arranged. That allowed the so-called Beignet bonds to trade in line with Meta’s corporate debt.
For those whose head is spinning with the circularity involved, this is how we described the deal last week when it was first floated:
Broadcom is backstopping a massive $36 billion private credit SPV with Apollo and Blackstone which will help Anthropic buy Google chips... made by Broadcom. Oh, and yes: Google owns 14% of Anthropic...
*BROADCOM: WORK WITH APOLLO, BLACKSTONE SERVES OPENAI, ANTHROPIC
Translation: Broadcom is backstopping a massive $36 billion private credit SPV with Apollo and Blackstone which will help Anthropic buy Google chips... made by Broadcom.
— zerohedge (@zerohedge) June 3, 2026
But wait, there's more... because if that wasn't enough, Morgan Stanley, which advised Broadcom and arranged the transaction, is also lending money to investors participating in the deal!
And just because this is a "chip-backed" off-balance sheet SPV where nobody really knows who holds the debt, the monstrous circularity of all the deal aspects will be ignored until the AI credit bubble cracks.
As for the punchline: demonstrating the insane frenzy of anything involving AI, investors involved in the deal did not even know what they were investing in! According to the FT, investors pitched on the deal were not given early access to Anthropic’s financials ahead of its IPO.
Not everyone involved in the deal is a total idiot: some investors passed on the deal over the delayed-draw format of the debt, which drives down yields because the money can be withdrawn in multiple tranches over a period of time.
Yet despite the smashing success of the deal, one glaring question remains. Recall, last week SpaceX penned a massive deal with Google (to urgently burnish the IPO candidate's financials just days ahead of its IPO), according to which Google will pay Elon Musk $920 million a month for access to about 110,000 Nvidia GPUs (unlike its hyperscaler peers, SpaceX has plenty of spare compute to rent out). And yet, despite seemingly telegraphing it is dramatically "compute constrained" as the SpaceX deal implies, it still has plenty of chip available that it can sell $35B of their chips to their biggest competitor, Anthropic.
This wasn't the only such deal: just days prior, Anthropic (which will use proceeds from this private credit SPV to purchase Google chips made by Broadcom), agreed to pay $1.5 billion a month for access to 325,000 Nvidia GPU also held by SpaceX. No wonder these sham agreements were structured so they can be terminated by either party after December 2026.
For those shaking their heads at these glaring examples of circular bubble euphoria, fear not: you will have plenty of opportunities to enjoy more such deals (going back to our point up top): Broadcom chief executive Hock Tan said the company hoped to connect “investor partners with the strongest balance sheets to deliver at scale sufficient compute capacity at the lowest cost”, pointing to the deal with Apollo and Blackstone as the first of many transactions to come.
Tyler Durden
Mon, 06/08/2026 - 23:06
Four leading AI models discuss this article
"The reliance on Broadcom's residual value support to manufacture investment-grade ratings on volatile AI hardware creates a systemic 'circularity trap' that obscures the true credit risk of the underlying AI startups."
This $35 billion 'Project Big Sky' deal is a masterclass in synthetic credit engineering, designed to bypass traditional bank constraints by shifting risk into private credit SPVs. By using Broadcom as a guarantor, Apollo (APO) and Blackstone (BX) have effectively repackaged venture-stage AI risk into investment-grade-lite paper, feeding the insatiable appetite of insurance vehicles like Athene. This circularity—where Google, Broadcom, and Anthropic are essentially financing each other’s growth—creates a feedback loop that masks true demand. While this structure lowers immediate borrowing costs, it creates a 'cliff risk' if GPU depreciation accelerates faster than the lease payments, potentially leaving credit holders holding rapidly obsolescing silicon.
If AI compute remains the primary bottleneck for global GDP growth, these chips function as 'digital oil' with predictable, long-term utility, making the credit risk fundamentally safer than traditional corporate debt.
"The deal is structurally sound but transfers AI capex ROI risk from balance sheets to off-balance-sheet SPVs, concentrating tail risk in Broadcom's guarantee and private credit funds' ability to absorb losses if utilization disappoints."
This article conflates three separate concerns—SPV structuring, circular deal-making, and AI capex financing—into one bubble narrative, but the mechanics don't support the apocalyptic framing. Yes, $35B in off-balance-sheet chip financing is large. Yes, Broadcom's residual-value guarantee creates moral hazard. But the deal priced rationally: A1/A2 tranches at 1% and 5.75% over Treasuries reflect genuine credit quality; junior debt at 8.5% prices tail risk. The real issue isn't the structure—it's whether Anthropic's lease payments sustain. The article's circularity complaint (Google owns Anthropic, funds chips, Broadcom backs the debt) is overblown; this is vertical integration, not fraud. What *should* worry you: if AI capex ROI disappoints, lease cash flows evaporate, and Broadcom's guarantee becomes the only backstop for $30B in senior debt.
If Anthropic's chip utilization and revenue scale as planned, lease payments flow reliably, and Broadcom's guarantee never gets tested, this deal simply accelerates AI infrastructure deployment without systemic risk—making the bubble narrative premature.
"APO and BX are warehousing concentrated AI credit risk in opaque SPVs that the article understates by focusing on size rather than the junior tranche's direct exposure to untested lessee cash flows."
The $35B 'Big Sky' SPV lets APO and BX originate high-yield junior debt and equity while Broadcom backstops senior tranches, but the structure hides Anthropic's unshared financials from most investors and ties repayment to chips whose residual value rests on unproven long-term demand. With $400B+ in AI debt already slated through 2028, any slowdown in lease payments or faster GPU depreciation would first hit the $4.5B junior piece and Athene's A2 holdings. The delayed-draw feature and post-Dec-2026 termination clauses further compress yields while leaving originators exposed to rapid obsolescence.
Broadcom's investment-grade residual guarantee and the senior notes' mid-IG private ratings may cap losses far below headline size, allowing APO and BX to earn attractive risk-adjusted spreads on what is effectively sponsored paper.
"Off-balance-sheet SPVs tied to AI hardware create leverage and opacity; a material AI demand downturn or sharper depreciation of GPUs could devastate junior-tranche losses even with Broadcom's backstop."
This deal signals continued appetite for AI infrastructure funding, with Apollo/Blackstone wiring a $35B private credit SPV to back Anthropic’s chip lease model. A1 $6B at ~1% over Treasuries, A2 $24B at ~5.75%, and $4.5B in junior debt at 8.5% — plus $800M equity from Atlas SP Partners — are backed by Broadcom's residual-value support. Yet the structure relies on long-duration lease cash flows and the uncertain long-term value of GPUs, while the junior tranche bears outsized risk. The hype around ‘hundreds of billions’ of SPVs glosses over counterparty, liquidity, and mark-to-market risk in a highly opaque sector.
If AI capex slows or GPU values deteriorate faster than modeled, the junior debt could implode despite Broadcom’s backstop; the deal’s complexity also masks true leverage and liquidity risk.
"The deal's circular financing structure creates a synthetic liquidity trap that masks true demand and risks massive balance sheet impairment if AI ROI fails."
Claude, you’re dismissing the 'circularity' too easily. This isn't just vertical integration; it’s a synthetic liquidity trap. By using Broadcom to backstop the SPV, Google and its ecosystem are effectively lending to themselves to inflate Anthropic’s valuation. If the underlying GPU utility fails to generate organic, non-Google-funded revenue, this entire structure collapses into a massive balance sheet impairment for the guarantors. The 'moral hazard' isn't a feature—it's the primary systemic vulnerability.
"The systemic risk isn't self-dealing; it's correlated lease-payment failure across the entire $400B AI debt stack if capex ROI disappoints."
Gemini conflates two distinct risks. Yes, Google funding Anthropic creates moral hazard—but that's a valuation problem, not a *structure* problem. The real cliff: if Anthropic's lease payments fail, Broadcom's guarantee only covers residual GPU value, not the full $30B senior debt. That gap is the systemic vulnerability. The circularity argument assumes Google *wants* Anthropic to fail; actually, Google's incentive is to keep it solvent. The actual risk is external—AI capex ROI disappointing industry-wide, forcing all lessees to renegotiate simultaneously.
"Simultaneous lease exits post-2026 could trigger sector contagion beyond isolated ROI failures."
Claude understates contagion from simultaneous lease renegotiations. The post-Dec-2026 termination clauses allow multiple AI firms to exit if ROI disappoints, hitting junior debt first but pressuring A2 tranches via liquidity evaporation. With $400B+ AI debt pipeline, Athene's holdings face mark-to-market hits from utilization opacity, turning isolated capex risk into sector-wide credit tightening that Broadcom's residual guarantee cannot fully mitigate.
"The real risk is correlated stress across the AI capex cycle that could trigger simultaneous lease renegotiations after 2026, overwhelming Broadcom's residual backstop and depressing both Athene's and senior note values."
Grok raises a contagion concern, but the bigger flaw is correlation risk across the entire AI capex cycle. Termination clauses after 2026 could trigger a wave of simultaneous renegotiations if ROI stalls; that would hit the whole stack, not just juniors. Broadcom's residual backstop may be insufficient if GPU residual values deteriorate quickly and lease cash flows collapse, leaving Athene and senior notes exposed to mark-to-market losses. The assumption of isolated stress is too optimistic.
The panel generally agrees that the $35B 'Project Big Sky' deal, while innovative, carries significant risks. The deal's structure, which relies on synthetic credit engineering and off-balance-sheet financing, creates potential vulnerabilities such as cliff risk, moral hazard, and systemic vulnerabilities in case of a slowdown in AI capex ROI or lease payment defaults.
None explicitly stated by the panel.
Rapid GPU depreciation leading to lease cash flow evaporation and a collapse of the SPV structure, leaving credit holders with obsolescing silicon and potential mark-to-market losses.