Nvidia Officially Confirms $105 Billion Commitment to Back SoftBank's Ohio Data Center
Nashnova编辑部
Jensen Huang confirmed Nvidia will provide ~$105 billion in credit support and a $1.5 billion equity stake for SoftBank's SB Energy data center campus in Ohio — a move that pushes Nvidia from chip supplier into deep financial backer of downstream AI infrastructure.
What does this deal actually look like?
Nvidia will provide ~$105 billion in credit support for SB Energy's data center campus in Pike County, Ohio, plus a $1.5 billion direct equity investment.
Nvidia's guarantee covers only specific portions of lease and power payments plus residual-value commitments — not the full cost of the campus. This means → Nvidia's exposure is capped and conditional, not an open-ended backstop.
The guarantees phase in as the campus comes online between 2028 and 2030, then step down as OpenAI pays rent and capacity ramps. In plain terms = the more reliably the tenant pays, the lighter Nvidia's burden becomes.
How big is the campus, and what could Nvidia earn?
The campus is planned for up to 10 GW total capacity. Nvidia's initial commitment covers 4.25 GW; whether it backs the remaining 3.75 GW is undecided.
OpenAI is the anchor tenant on a 20-year lease. Nvidia is the sole supplier of AI compute for the site. This means → as long as the campus runs, every chip inside is Nvidia's.
Huang estimated each generation of AI factory systems requires roughly 1.5 million Nvidia GPUs, with potential hardware revenue of $150–200 billion.
Where does the power come from?
SB Energy plans to build 10 GW of new generation capacity, including 9.2 GW of natural-gas power, to support up to 8 GW of IT load.
SB Energy and AEP Ohio plan to invest at least $4.2 billion in regional grid infrastructure. This reflects a shift: the real bottleneck for large-scale AI data centers is no longer chip supply — it is power.
Why did the commitment shrink from $250 billion to $105 billion?
Nvidia had considered backing up to $250 billion but scaled back to $105 billion after investors flagged concerns about risk exposure. The equity stake also dropped from $3 billion to $1.5 billion.
The financing structure resembles Meta's off-balance-sheet "Project Beignet" model. In plain terms = chip giants are no longer just selling hardware — they are starting to act like banks, guaranteeing and financing their customers' infrastructure.
The cut exceeded half the original figure. This reflects that the market is still actively repricing the risk of a chip company carrying hundreds of billions in guarantees.
What is the "circular financing" concern?
The core worry: Nvidia channels profits to downstream AI companies, which then spend that money buying Nvidia chips — the cash loops back to Nvidia.
Huang responded that Nvidia is using its scale and long-term visibility into customer demand to lock in critical infrastructure, consistent with supply-chain management logic — not a financial round-trip.
He added that top AI labs are growing faster than their own balance sheets can support. In plain terms = these companies are expanding too fast to borrow enough on their own, so they need Nvidia to step in as guarantor.
When will we know if this deal actually works?
The first real test is whether initial capacity comes online on schedule in 2028.
That is also the first observable window for the market to assess Nvidia's actual risk exposure. This means → until then, the $105 billion commitment is largely on paper; 2028 is when the real stress test begins.
Content is for reference only, not financial advice.