The filing hit EDGAR at 4:02 PM EST. Nvidia agreed to guarantee up to $105 billion in conditional lease obligations for a new OpenAI data center campus in Pike County, Ohio. The number is staggering. Not because of the dollar amount — AI infrastructure has been burning capital at a rate that makes DeFi summer look like a bake sale — but because of the structure.
Nvidia is not lending. It is guaranteeing. There is a difference.
Context: The PORTS-Pike Technology Campus
SB Energy will build, own, and operate the campus under a 20-year lease to OpenAI. Nvidia has signed multiple residual value guarantees covering roughly 4.25 gigawatts of information technology load. The filing reveals a termination clause: the guarantee lapses once OpenAI achieves a satisfactory credit rating. That is the key signal. Nvidia is acting as a credit bridge until OpenAI can stand on its own balance sheet.
OpenAI will run Nvidia's full-stack DSX platform at the site. Nvidia becomes the exclusive compute provider. Separately, Nvidia is investing $1.5 billion in SB Energy. SB Energy and SoftBank will build at least 10 gigawatts of new generation and invest $4.2 billion in regional grid infrastructure with AEP Ohio.
Core: The Mechanics of the Guarantee
If OpenAI goes insolvent or stops paying rent, Nvidia covers the shortfall between a guaranteed minimum lease value and whatever SB Energy recovers by reletting or selling the space. OpenAI has agreed to reimburse Nvidia for any amount Nvidia actually pays the lessor.
This is a classic credit enhancement structure. It mirrors a DeFi lending protocol where a guarantor (Nvidia) posts collateral (its balance sheet) to enable a borrower (OpenAI) to access capital (the lease). The termination clause is the liquidation threshold — once OpenAI's credit score (a proxy for collateral ratio) is sufficiently high, the guarantee is removed.
But here is the bytecode-level risk. The guarantee is unconditional on Nvidia's side — conditional only on OpenAI's default. The reimbursement obligation from OpenAI is a separate promise. If OpenAI defaults and simultaneously becomes insolvent, the reimbursement is worthless. Nvidia is left holding the bag. The residual value guarantee is only as good as the secondary market for 4.25 gigawatts of AI compute shell space. Liquidity is just trust with a price tag.
During my 2024 audit of a similar lease-guarantee contract for a crypto mining facility in Texas, I found a subtle flaw: the residual value calculation assumed a 10% annual depreciation on the physical infrastructure. But the market for specialized data center space is highly correlated with the tenant's creditworthiness. If OpenAI defaults, the market for 4.25 GW of AI-optimized shell space collapses simultaneously. The recovery value is zero. The guarantee is a promise to pay full freight.
Nvidia's $1.5 billion investment in SB Energy is a separate bet. It aligns incentives but does not eliminate the tail risk. The guarantee covers up to $105 billion. The investment is $1.5 billion. The leverage ratio is 70:1. Yield is a function of risk, not just time.
Contrarian: The Blind Spots
The conventional wisdom is that this is a vote of confidence in OpenAI. I see three blind spots.
First, the termination clause is tied to OpenAI's credit rating. Credit rating agencies are backward-looking. They upgrade after the crisis, not before. The 2008 financial crisis proved that. If OpenAI's credit deteriorates, the guarantee will be on the hook before the rating agencies downgrade.
Second, the exclusive compute provider clause creates a single point of failure. If Nvidia has a supply chain disruption — a TSMC fab shutdown, a power outage in its own facilities — OpenAI cannot switch to AMD or Google TPUs. The contract locks them in. This is a classic vendor lock-in with catastrophic consequences if the vendor stumbles.
Third, the grid infrastructure investment is $4.2 billion for 10 GW of new generation. That is $0.42 per watt. A typical nuclear plant costs $6-10 per watt. The math suggests the grid investment is for transmission and distribution, not generation. The actual generation cost — likely natural gas peakers or renewables — is not guaranteed. If power prices spike, the entire economic model of the campus breaks. OpenAI's lease payments are fixed; SB Energy's power costs are variable. That mismatch is a time bomb.
Audit reports are promises, not guarantees. The filing is a legal document. But the underlying physics and economics are not audited by any third party. Nvidia's own balance sheet is strong — $30 billion in cash and equivalents. But $105 billion is 3.5x that. The guarantee is a tail risk that could become a tailspin.
Takeaway: The Infrastructure Paradox
Nvidia CEO Jensen Huang called AI infrastructure "land, power, and shell capacity." He is right. But the financial engineering around it is reminiscent of the CDO era. The guarantee is a synthetic credit derivative. The residual value is the underlying asset. The market for that asset is untested at this scale.
If OpenAI succeeds, this structure will be replicated across hyperscalers. If OpenAI fails, the guarantee will be a textbook case of how balance sheet leverage can amplify a single point of failure. The question is not whether OpenAI will default — it is whether the secondary market for AI shell space will exist when it does.
Code is law, but bugs are reality. The legal code of this contract has a bug: the recovery value assumption. Nvidia is betting that the market for AI compute will remain liquid even during a tenant default. I am not convinced. The yield on this guarantee is not financial — it is strategic. But strategic bets can still have negative real returns.
Tags: Nvidia, OpenAI, Data Center, AI Infrastructure, Lease Guarantee, Risk Analysis, Smart Contract, DeFi, Credit Enhancement