Nvidia Commits $105 Billion Guarantee to Back OpenAI's Ohio Infrastructure

A New Calculus for AI Infrastructure

Nvidia has structured a guarantee arrangement worth up to $105 billion to support OpenAI's lease of a data center complex in Ohio, according to the chipmaker. The facility is being developed by SB Energy, a portfolio company backed by SoftBank. The guarantee represents one of the largest infrastructure financing commitments from a semiconductor company to date.

The arrangement was disclosed alongside a separate $1.5 billion equity investment Nvidia is making directly into SB Energy. That investment follows an earlier $1 billion commitment from OpenAI and SoftBank to expand SB Energy's data center footprint.

At DailyTechWire, we've tracked the convergence of chip design, power infrastructure, and AI model development across Asia and North America for the past eighteen months. This deal crystallizes a trend: the companies building inference and training hardware are no longer content to sell chips and walk away. They are embedding themselves in the capital stack of the facilities those chips will occupy.

Why Hardware Vendors Are Writing Checks for Buildings

The guarantee structure is notable because it shifts risk. Traditionally, a data center developer like SB Energy would secure financing through project debt, leases, or equity from infrastructure funds. The tenant - in this case, OpenAI - would sign a long-term lease, and the landlord would use that contract to raise capital.

Nvidia's guarantee changes the equation. By backstopping the lease obligation, Nvidia effectively reduces the financing cost for SB Energy and makes it easier for OpenAI to lock in capacity without tying up its own balance sheet. The chipmaker benefits by ensuring that a marquee customer has access to the physical infrastructure required to deploy Nvidia's H100 and next-generation Blackwell chips at scale.

This is not philanthropy. Nvidia's revenue growth depends on customers having somewhere to install its accelerators. If data center capacity becomes a bottleneck - due to power constraints, permitting delays, or capital scarcity - chip sales stall. By guaranteeing lease payments, Nvidia is buying optionality on future hardware orders.

The Ohio Site and SB Energy's Ambitions

SB Energy has been assembling sites across the United States with access to grid-scale power and fiber connectivity. Ohio offers both: legacy coal infrastructure that can be repurposed for high-voltage transmission and proximity to Midwest fiber routes linking Chicago, Columbus, and Pittsburgh.

The SoftBank-backed entity has positioned itself as a turnkey provider for hyperscale AI tenants, offering not just rack space but integrated power contracts and cooling systems designed for the thermal loads of dense GPU clusters. The $1 billion investment from OpenAI and SoftBank earlier this year was earmarked for site preparation, substation upgrades, and securing long-term power purchase agreements with regional utilities.

Nvidia's additional $1.5 billion equity stake gives the chipmaker a seat at the table during site planning. That influence matters: decisions about rack layout, cooling architecture, and network topology directly affect chip performance and longevity. By investing in SB Energy, Nvidia can shape those choices before the first server arrives.

Capital Allocation in the Age of Inference

The deal also reflects a broader shift in how AI companies allocate capital. Training large language models remains expensive, but inference - running those models at scale for millions of users - is where the long-term compute demand lives. Inference workloads require sustained capacity, not burst performance, and they favor geographic distribution to reduce latency.

OpenAI's decision to lease rather than build gives the company flexibility. If model architectures change or if a new generation of chips offers better performance per watt, OpenAI can renegotiate or relocate without being stuck with stranded assets. The trade-off is that leasing ties up ongoing cash flow. Nvidia's guarantee mitigates that concern by lowering the effective cost of the lease, likely through reduced interest rates or more favorable amortization terms.

For Nvidia, the $105 billion figure is a ceiling, not a check. The guarantee is contingent on OpenAI's lease obligations over time, which could stretch across a decade or more. If OpenAI meets its payments, Nvidia pays nothing. The risk is that OpenAI defaults or renegotiates, in which case Nvidia would need to cover the shortfall or take ownership of the lease. Given OpenAI's revenue trajectory and Microsoft's backing, that scenario is remote - but the exposure is real.

Regional Implications and the Race for Edge Capacity

From an Asia-forward lens, the Ohio deal is a signal of how U.S. AI infrastructure is consolidating around a handful of anchor tenants. OpenAI, Microsoft, Google, and Meta are pre-leasing capacity at a pace that leaves smaller model developers scrambling for access. That dynamic is pushing frontier AI work toward regions with surplus power and available land: the U.S. Midwest, parts of Scandinavia, and increasingly, Malaysia and Indonesia.

SoftBank's involvement also matters. The conglomerate has made data center infrastructure a priority across its Vision Fund portfolio, with investments in Southeast Asian colocation providers and renewable energy projects. The SB Energy model - bundling power, cooling, and connectivity under one entity - is being tested in Japan and South Korea as well. If the Ohio site proves that Nvidia-backed guarantees can de-risk large leases, expect similar structures to appear in Osaka, Busan, and Johor Bahru.

What This Means for the Next Wave of AI Buildout

Nvidia's willingness to guarantee $105 billion in lease obligations is a bet that demand for AI compute will remain inelastic even as chip prices decline. It also suggests that the company views infrastructure financing as a competitive moat. Rivals like AMD and Intel do not yet have the cash flow or the customer relationships to offer comparable guarantees.

The arrangement sets a precedent. If chip vendors are willing to underwrite data center leases, they can effectively lock in customers for the duration of those agreements. That reduces churn and makes it harder for competitors to displace incumbents, even if their hardware is marginally better or cheaper.

For OpenAI, the deal provides certainty at a time when inference demand is outpacing its ability to provision capacity. The Ohio site will likely house a mix of training clusters and inference endpoints, supporting both research workloads and production traffic for ChatGPT and API customers. With Nvidia's guarantee in place, OpenAI can scale without the balance-sheet drag of long-term lease liabilities.

The broader question is whether this model is sustainable. If multiple chip vendors start offering guarantees to secure customer loyalty, the practice could inflate data center valuations and create hidden leverage in the AI supply chain. For now, Nvidia has the revenue and the relationships to make the math work. Whether others can follow - or whether regulators take notice - remains to be seen.