NewsFlash Articles Data Fundraising BlockBeats Pro

Huang Renxun, Ultraman, and Son Jeonghyeon "Form Alliance for 20 Years"

Read this article in 18 Minutes

OpenAI, NVIDIA, and SoftBank's SB Energy have confirmed the construction of a large-scale data center in the United States. OpenAI has signed a 20-year lease, with NVIDIA providing a maximum of $105 billion in guarantees.

Original Title: "Elon Musk, Ultraman, and Masayoshi Son: 20 Years of Collaboration"

OpenAI is securing its computing power needs for the next 20 years, while NVIDIA is expanding beyond chips.

On August 17, local time in the U.S., NVIDIA, OpenAI, and SB Energy under SoftBank confirmed a collaboration to build a large-scale data center in the U.S. SB Energy will be responsible for development and operation, OpenAI will be the customer, and the data center will fully utilize NVIDIA's computing power.

The role played by NVIDIA in this tripartite transaction has attracted the most attention. It is not only a computing power supplier but also provides a maximum $105 billion guarantee for OpenAI's long-term lease and directly invests in the project's construction. What was originally a transaction between developers, tenants, and financiers now includes a chip company.

Why is NVIDIA taking on such risk, and why is OpenAI locking in a 20-year computing power lease?

One fact and trend is this: under the expanding demand for computing power, land, electricity, and data center capacity have become bottleneck resources, sparking competition among various parties. However, in the view of Elon Musk, the balance sheet of cutting-edge AI labs falls short of cloud giants, leading to the emergence of a "chip company participating in guarantee" approach.

In reality, though, the cloud giants are also "struggling under the weight."

According to The Wall Street Journal's analysis of the latest financial filings of 9 large tech companies, these companies' off-balance-sheet commitments related to AI (not included in the balance sheet, mainly from forward payment obligations for chip purchases and long-term data center leases) total nearly $3 trillion, three times the sum of current lease liabilities and long-term debts.

10GW Lease with NVIDIA Guarantee

The data center project involving OpenAI, SB Energy under SoftBank, and NVIDIA is located in the PORTS-Pike area of Ohio, USA. It plans to build at least 10GW of new energy generation capacity. Once completed, the area will exclusively deploy NVIDIA's computing infrastructure, ultimately creating approximately 8GW of AI factory capacity.

The project will be constructed in two phases. The first phase is planned to have a capacity of 4.25GW, with the initial 800MW expected to be operational by 2028, mainly utilizing the existing AEP Ohio infrastructure. Subsequently, the project will need to continue building power plants, transmission lines, and other grid facilities, and then expand the data center capacity to the intended target.

NVIDIA may also secure an additional 3.75GW of capacity based on future needs, but there is currently uncertainty in infrastructure, permitting, and other aspects. NVIDIA is not obligated to lease all of this capacity.

Construction began on the Southern Ohio Data Center Park in March of this year

SB Energy and SoftBank plan to invest at least $4.2 billion in building new regional grid infrastructure. SB Energy is responsible for constructing, owning, and operating the data centers, with OpenAI using the capacity once it is built and delivered.

Currently, OpenAI has signed a 20-year lease with SB Energy. The agreement states that OpenAI will only start paying rent once the corresponding capacity is completed and available for lease.

Signing a long-term capacity agreement in advance allows OpenAI to secure its future computing power needs. However, this does not mean that NVIDIA will have to pay OpenAI's full rent for 20 years.

According to the agreement, NVIDIA's maximum payment obligation (or total payment cap) for this transaction is $105 billion, primarily for costs related to land, power, data center infrastructure, and other expenses.

NVIDIA has adopted a residual value guarantee structure: if OpenAI stops leasing, SB Energy will need to find a new tenant. If no new lease is secured, the assets may be considered for sale. Only when the above methods cannot cover the agreed minimum value, will NVIDIA make up the difference.

Therefore, the $105 billion guarantee corresponds to the remaining value of the completed data center assets and will be phased in as the project is built and put into operation, roughly covering the period from 2028 to 2030. As OpenAI pays rent and the data center capacity comes online gradually, NVIDIA's actual risk exposure will also decrease progressively.

NVIDIA is willing to do this precisely because of this risk exposure. Even if OpenAI reduces its future usage, the built computing capacity can still be transferred to cloud service providers, enterprises, AI labs, and startups.

Early "Land Grab" for Land and Power

On the same day this news was released, Huang Renxun wrote an article explaining why NVIDIA is participating in such a project. His assessment is that AI factories require more and more things. In the past, advanced chips, packaging, memory, and networking were the main investments in AI infrastructure. Now, access to land, power, and data centers must also be secured in advance.

In Huang Renxun's view, large cloud service providers and investment-grade enterprises usually have a sufficiently large balance sheet and the ability to sign long-term contracts and build their own infrastructure. However, cutting-edge AI labs may not necessarily have these conditions.

These companies' training and inference needs are growing rapidly, and their revenue may also increase accordingly. But to secure decades of land, power, and data centers in advance, stable cash flow and strong enough financing capabilities are required, which many AI labs currently lack.

As a result, a new bottleneck has emerged.

Huang Renxun wrote that the growth of these companies is "not limited by algorithms or customer demand, but by the availability of computing power."

NVIDIA's involvement in Data Center Infrastructure (LPS, Land, Power, and Shell) aims to address this issue. However, this does not mean that NVIDIA is preparing to provide similar services to all customers. Huang Renxun emphasized that NVIDIA will only select a few high-quality sites with clear, long-term computing needs.

Huang Renxun revealed that each generation of NVIDIA AI Factory systems deployed in the PORTS-Pike campus may correspond to around 1.5 million NVIDIA GPUs, or approximately $150 billion to $200 billion in NVIDIA revenue.

The term "each generation" is crucial here. For NVIDIA, over a 20-year agreement period, what is actually being locked in is the long-term infrastructure that will host its computing systems, rather than a fixed order for a particular generation of GPUs.

OpenAI's long-term commitment further amplifies this opportunity.

Huang Renxun stated that OpenAI's existing and planned commitments correspond to approximately 12GW of NVIDIA computing power. If PORTS-Pike continues to expand, the related capacity will increase. At this scale, by 2030, OpenAI's deployment opportunities are estimated to correspond to around $600 billion in NVIDIA computing power value.

In the past, chip companies "invested" in customers, sparking discussions about circular financing. Now, the relationship between the two has gone further, with chip companies directly entering the data center space to underpin data center construction. On one hand, this provides a "guaranteed endorsement" for the computing power needs of cutting-edge labs, and on the other hand, data center construction will bring continuous orders to themselves.

A $3 Trillion Off-Balance-Sheet Commitment

The computing power story of PORTS-Pike involves not only NVIDIA and OpenAI.

Over the past two years, both AI companies and large tech firms have been frantically building data centers. However, an increasing amount of this infrastructure is not being directly owned by them, but rather secured through leases, long-term procurement agreements, joint ventures, and other financing structures.

The Wall Street Journal recently analyzed the latest filings from 9 large tech companies, including Alphabet, Amazon, Microsoft, Meta, Oracle, NVIDIA, Broadcom, SpaceX, and AMD, and found that as of the most recent disclosure, these companies' off-balance-sheet AI-related commitments totaled around $3 trillion.

Compared to the approximately $600 billion in capital expenditures over the past year, the scale of off-balance-sheet commitments signed by these companies is much larger.

Meta's Hyperion data center is a typical example.

This data center in Louisiana covers an area equivalent to 1700 football fields and is being built by a joint venture owned by a fund managed by Blue Owl Capital. Meta is a minority partner and a tenant, with its rent providing cash flow to bondholders.

Before starting to pay rent, this obligation will not be fully reflected on Meta's balance sheet. As of June this year, Meta's disclosed but as-yet-unstarted lease obligations amounted to $347 billion, including the Hyperion project.

As of June, Meta has leased the Hyperion data center in Louisiana, but rent payments have not commenced, and the related lease obligations have not been entirely included on the balance sheet

According to data, the total commitments for lease payments yet to begin for the 9 companies amount to around $1.2 trillion, about 4 times the disclosed amount from a year ago. Purchase commitments and other contractual obligations amount to around $1.9 trillion.

Among these, Alphabet's changes are particularly noticeable.

As of June 30, Alphabet's purchase commitments and contractual obligations reached $811 billion, up from $332 billion three months earlier. Alphabet explained that these obligations mainly involve "technology infrastructure and inventory," as well as agreements to ensure data center energy supply. Some of the energy agreements even extend to 2054. However, Alphabet did not provide a detailed explanation as to why these commitments increased by nearly $480 billion in just one quarter.

The expansion risk involves more than just data center leasing and chip procurement. Some companies' commitments also include buying back shares of other companies or providing guarantees for other tenants' leases. NVIDIA itself has committed to a $27 billion equity investment between April 26, 2026, and the end of the 2027 fiscal year.

However, another risk lies in debt expansion. Some tech companies have begun to frequently access the capital markets for debt financing. In the recent earnings reports of Alphabet and Amazon, free cash flow has been negative, and capital expenditures have exceeded cash generated from operating activities.

Alphabet, Amazon, and Meta's originally healthy cash flow (blue bars) are expected to collectively turn negative in the remaining time of 2026 and in 2027 (gray bars)

These figures have not yet fully reflected the cash flow pressure that may arise from future off-balance-sheet commitments amounting to trillions of dollars. What's more troubling is that many purchase commitments and long-term leases cannot be easily canceled. In other words, even if future AI demand does not meet expectations, companies will have to foot the bill for the signed agreements. In such a scenario, tech giants will be forced to cut other expenses and may further borrow money to maintain this infrastructure.

In an April report, Morgan Stanley's accounting analysts warned that as such off-balance-sheet commitments become more frequent, larger in scale, and more complex in structure, assessing a company's true leverage level will become increasingly difficult for investors.

Now, PORTS-Pike stands at the forefront of this trend. But who can guarantee that the computing power demand will continue to expand aggressively without slowing down?