Nvidia’s $105B Guarantee: The Hidden Leverage Play on AI Compute

Nvidia just put $105 billion on the line for OpenAI's Ohio data center. That's not a loan. It's a residual value guarantee covering 4.25 gigawatts of compute capacity, with an option for another 3.75 GW. Speed is the only currency that doesn't inflate. The filing dropped quietly. Nvidia will back the leases for the PORTS-Pike Technology Campus in Pike County, Ohio, built by SB Energy and leased to OpenAI for 20 years. If OpenAI stops paying rent or goes insolvent, Nvidia covers the shortfall between the guaranteed minimum lease value and whatever SB Energy recovers from reletting or selling the space. OpenAI has agreed to reimburse Nvidia for any actual payments. The guarantee terminates once OpenAI achieves a satisfactory credit rating. That termination clause is the key. This isn't charity. It's a bridge. OpenAI's credit isn't strong enough to lock in 20-year leases on 4.25 GW of compute capacity alone. Nvidia steps in as the guarantor, effectively lending its own balance sheet to the deal. Once OpenAI's credit rating improves—likely after a few years of revenue growth or an IPO—the guarantee vanishes. Speed is the only currency that doesn't inflate. But why does Nvidia care? Because the campus will run Nvidia's full-stack DSX platform exclusively. Nvidia becomes the sole compute provider for that site. CEO Jensen Huang called AI infrastructure 'land, power, and shell capacity.' He's securing long-lived infrastructure for his own chips. The $1.5 billion Nvidia invested separately in SB Energy is a down payment on that exclusivity. From a structural perspective, this is a synthetic lease guarantee with a call option on future compute demand. The initial 4.25 GW of information technology load is the floor. The optional 3.75 GW is Nvidia's discretion. Capacity comes online in phases starting 2028. That timeline matters. By 2028, the AI hardware cycle will have turned at least twice. The guarantee protects Nvidia against the risk that OpenAI's demand collapses before the campus is fully built. But if demand holds, Nvidia locks in a massive captive customer for its flagship hardware. I've seen similar structures in crypto lending—residual value guarantees on collateralized debt positions. The math is the same. The guarantor takes the downside risk in exchange for exclusive upside. Nvidia is betting that OpenAI's compute needs will grow faster than the market expects. If it's wrong, the $105 billion guarantee becomes a real liability. If it's right, the guarantee never triggers, and Nvidia captures the full margin on hardware sales for the next decade. The contrarian angle is that this deal exposes Nvidia to a single counterparty risk. OpenAI is not profitable. Its revenue is growing, but its cost structure is ballooning. The guarantee is a form of synthetic leverage. Nvidia is effectively underwriting OpenAI's credit risk without being compensated through interest—only through hardware exclusivity. That's a bet on the durability of the AI narrative. What if the AI hype cycle peaks before 2028? The 4.25 GW of compute capacity could become a stranded asset. SB Energy would have to relet the space to other hyperscalers, who might not need Nvidia hardware. The residual value would drop. Nvidia would be on the hook. Meanwhile, the energy grid in Ohio is under strain. SB Energy and SoftBank are building at least 10 GW of new generation and investing $4.2 billion in grid infrastructure with AEP Ohio. That's a separate risk. If the grid upgrades are delayed, the campus doesn't go live, and the guarantee timeline shifts. Regulatory scrutiny is another factor. The Federal Energy Regulatory Commission and state regulators are watching large-scale power purchases for AI data centers. Any policy change could affect the economics. But the market is ignoring these risks. The narrative is that Nvidia is securing the future of AI. The reality is that Nvidia is leveraging its own balance sheet to manufacture demand for its own products. Speed is the only currency that doesn't inflate. What to watch next: OpenAI's credit rating. Moody's or S&P will likely upgrade it within 18 months if revenue growth continues. That would trigger the termination clause and remove the liability from Nvidia's books. Also watch the construction timeline. Any delays beyond 2028 will compress the cash flow window for SB Energy. Finally, watch Nvidia's own debt-to-equity ratio. The $105 billion guarantee is off-balance-sheet, but analysts will start asking about contingent liabilities. This is the template for the next phase of AI infrastructure. Guarantees, not direct loans. Exclusivity, not open market. The hyperscalers are building their own fiefdoms. Nvidia is the landlord who also owns the keys to the servers.