TechFlow Logo
Login/ Sign up
ETH Gas
Gwei
Fear
gas
Jensen Huang, Sam Altman, Masayoshi Son "Stick Together for 20 Years"

Jensen Huang, Sam Altman, Masayoshi Son "Stick Together for 20 Years"

2026.08.18
Share
OpenAI, NVIDIA, and SoftBank's SB Energy confirm building large-scale data centers in the US, with OpenAI signing a 20-year lease and NVIDIA providing a guarantee of up to $105 billion.

Author: Su Yang, Tencent Technology

OpenAI is locking in computing power for the next 20 years in advance, and Nvidia is also reaching beyond chips.

On August 17 local time in the US, Nvidia, OpenAI, and SoftBank's SB Energy confirmed a partnership to build a large-scale data center in the United States.

Among them, SB Energy is responsible for development and operations, OpenAI will become the customer, and the data center will entirely adopt Nvidia computing power.

In the three-party transaction, the most attention-grabbing aspect is the role Nvidia plays.

It is both a computing power supplier and provides a bottom-line guarantee of up to $105 billion for OpenAI's long-term lease, while directly investing in project construction.

Originally a transaction between developers, tenants, and financing institutions, there is now an additional layer involving a chip company.

Why is Nvidia willing to bear such risks, and why has OpenAI locked in a 20-year computing power lease?

A fact and trend is: under the expansion of computing power demand, land, electricity, and data center capacity have become "bottleneck" resources, triggering competition and layout from all parties.

But in Jensen Huang's view, the balance sheets of frontier AI labs are not as strong as cloud giants, leading to the method of "chip company participation in guarantees."

But in reality, cloud giants are also "carrying a heavy load."

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

10GW Lease, Nvidia Provides Guarantee

The data center project resulting from this collaboration between OpenAI, SoftBank's SB Energy, and Nvidia is located in a park named PORTS-Pike in Ohio, United States.

It plans to build at least 10GW of new energy generation capacity. After the park is completed, it will exclusively deploy Nvidia computing infrastructure, ultimately forming about 8GW of AI factory capacity.

Project construction is divided into two phases.

The first phase plans a capacity of 4.25GW, of which the first batch of 800MW is expected to be put into use in 2028, mainly utilizing existing AEP Ohio infrastructure.

Afterwards, the project still needs to continue building power plants, transmission lines, and other grid facilities, and then expand data center capacity to the predetermined target.

Nvidia can also honor another 3.75GW of capacity guarantee based on future demand, but this part currently has uncertainties in terms of infrastructure, permit approvals, etc., and Nvidia is not obligated to lease all of it.

The southern Ohio data center park broke ground in March this year

SB Energy and SoftBank plan to invest at least $4.2 billion to build new regional grid infrastructure.

SB Energy is responsible for building, holding, and operating the data center, and OpenAI uses it according to the capacity that has been built and delivered.

Currently, OpenAI has signed a 20-year lease with SB Energy.

The agreement mentions that OpenAI will start paying rent only after the corresponding capacity is completed and available for lease.

Signing long-term capacity in advance allows OpenAI to lock in future computing power demand first.

But it does not mean Nvidia has to pay all the rent for 20 years on behalf of OpenAI.

According to the agreement, the maximum payment liability (or total payment obligation cap) Nvidia bears for this transaction is $105 billion, mainly targeting costs for infrastructure such as land, electricity, and data centers.

Nvidia is adopting a residual value guarantee structure: if OpenAI stops leasing, SB Energy must find a next tenant.

If there is no new lease, consider selling the relevant assets.

Only when the above methods still cannot cover the agreed minimum value will Nvidia make up the difference.

Therefore, the $105 billion guarantee corresponds to the residual value of the already built data center assets, and it will take effect in stages with project construction and put-in-use, roughly covering 2028 to 2030.

As OpenAI pays rent and data center capacity comes online successively, Nvidia's actual risk exposure will also gradually decrease.

The key to Nvidia's willingness to do this lies in this risk exposure.

Even if OpenAI reduces usage in the future, the already built computing capacity can still be transferred to cloud service providers, enterprises, AI labs, and startups.

Pre-emptively "Seizing" Land and Electricity

On the same day the above news was released, Jensen Huang wrote an article explaining why Nvidia would participate in such a project.

His judgment basis is that AI factories need more and more things.

In the past, advanced chips, packaging, memory, and networking were the main investments in AI infrastructure.

Now, land, electricity, and data centers must also be secured in advance.

In Jensen Huang's view, large cloud service providers and investment-grade enterprises usually have large enough balance sheets and the ability to sign long-term contracts and build infrastructure themselves.

But frontier AI labs do not necessarily possess such conditions.

The training and inference demands of these companies are growing rapidly, and revenue may increase accordingly, but to lock in land, electricity, and data centers for decades in advance, stable cash flow and strong enough financing capabilities are needed, which is exactly what many AI labs currently lack.

Thus, a new bottleneck emerged.

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

Nvidia's intervention in data center infrastructure (LPS, Land, Power and Shell) is to solve this problem.

However, this does not mean Nvidia is prepared to provide similar services to all customers.

Jensen Huang emphasized that Nvidia will only select a few high-quality sites with clear, long-term computing demands.

Jensen Huang revealed that each generation of Nvidia AI factory systems deployed in the PORTS-Pike park may correspond to about 1.5 million Nvidia GPUs, or about $150 billion to $200 billion in Nvidia revenue.

The "each generation" here is very important.

For Nvidia, within the 20-year agreement cycle, what is locked in is actually the infrastructure that supports its computing systems over the long term, not fixed orders for a certain generation of GPUs.

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

Jensen Huang stated that OpenAI's existing and planned commitments correspond to about 12GW of Nvidia computing capacity.

If PORTS-Pike continues to expand, the relevant capacity will also increase.

According to this scale, by 2030, OpenAI-related deployment opportunities will correspond to approximately $600 billion in Nvidia computing power value.

In the past, chip companies "took stakes" in customers, triggering discussions about circular financing.

Now the relationship between the two has gone a step further; chip companies are directly laying out data centers and providing a bottom-line guarantee for data center construction. On one hand, it provides "guarantee endorsement" for the computing power demands of frontier labs, and on the other hand, because data center construction will bring continuous orders to themselves.

$3 Trillion in Off-Balance Sheet Commitments

The computing power story of PORTS-Pike is not just about Nvidia and OpenAI.

In the past two years, AI companies and large tech companies have been frantically building data centers, but more and more infrastructure is not directly bought by them, but locked in advance through leases, long-term procurement agreements, joint ventures, and other financing structures.

The Wall Street Journal recently analyzed the latest securities files of 9 major tech companies including Alphabet, Amazon, Microsoft, Meta, Oracle, Nvidia, Broadcom, SpaceX, and AMD, and found that as of the most recent filing, these companies' off-balance sheet commitments mainly related to AI total about $3 trillion.

Compared to about $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 case.

This data center located in Louisiana covers an area equivalent to 1,700 football fields, and a joint venture held by a fund managed by Blue Owl Capital is responsible for construction.

Meta is a limited partner and also a tenant, and its rent provides cash flow for bondholders.

Before starting to pay rent, this part of the obligation will not be fully reflected on Meta's balance sheet.

As of June this year, the total lease obligations disclosed by Meta that have not yet started reached $347 billion, which includes the Hyperion project.

As of June, Meta has leased the Hyperion data center in Louisiana, but rent payments have not yet started, and the relevant lease obligations have not been fully included in the balance sheet

According to statistics, the lease payment commitments of the 9 companies that have not yet started total about $1.2 trillion, approximately 4 times the amount disclosed a year ago.

Purchase commitments and other contractual obligations reached about $1.9 trillion.

Among them, the change in Alphabet is particularly obvious.

As of June 30, Alphabet's purchase commitments and contractual obligations reached $811 billion, while three months ago it was $332 billion.

Alphabet explained that these obligations mainly involve "technology infrastructure and inventory," as well as agreements to ensure energy supply for data centers.

Some energy agreements even need to last until 2054.

But Alphabet did not explain in detail why relevant commitments increased by nearly $480 billion in just one quarter.

The risks of expansion are not only data center leases and chip procurement.

Some companies' commitments also include purchasing stock in other companies, or providing guarantees for leases of other tenants.

Nvidia itself has committed to making $27 billion in equity investments between April 26, 2026, and the end of the 2027 fiscal year.

But another risk lies in the expansion of debt.

Some tech companies have started to frequently enter the capital market for debt financing.

In the recently announced performance of Alphabet and Amazon, free cash flow is negative, and capital expenditures have exceeded cash generated by operating businesses.

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

And these figures do not fully reflect the cash flow pressure that trillions of dollars in off-balance sheet commitments may bring in the future.

What is more troublesome is that many purchase commitments and long-term leases cannot be easily canceled.

That is to say, even if future AI demand does not meet expectations, one must still pay for the signed agreements.

In this case, giants will be forced to cut other expenditures, and may also borrow further to maintain these infrastructures.

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

Now, PORTS-Pike stands at the forefront of this trend.

But who can guarantee that computing power demand will always expand infinitely aggressively and not slow down?