Foreign media report that the AI data center boom, while seemingly a race for computing power, is increasingly becoming a financing competition. Building a large-scale campus requires not only GPUs but also配套 land, grid connectivity, transformer equipment, cooling systems, fiber optics, and long-term power supply—individual projects may require investments of tens of billions of dollars.
By 2030, global AI infrastructure is estimated to require trillions of dollars in investment. Reuters previously noted that financing needs between 2026 and 2030 could reach $3.6 trillion. The article suggests that this means the entities bearing the risks of construction, lending, and holding long-term assets are often not the companies training the models themselves.
There are limited cash-rich individuals.
The article points out that large cloud providers like Microsoft can still directly invest cash using their own balance sheets, but many smaller AI companies and cloud service providers lack this capability. They often need to spend billions of dollars on GPUs and data center construction before generating stable revenue, and thus must rely on external capital.
Common practices include issuing corporate bonds and convertible bonds. In August 2026, Nebius announced a $4.5 billion convertible bond offering, with part of the proceeds allocated to its AI infrastructure business. The article suggests that such instruments allow equity investors to indirectly participate in data center expansion.
Data centers are being treated as infrastructure assets.
The article states that the financing methods for ultra-large AI campuses are increasingly resembling those of power plants, airports, or toll roads. Developers typically establish special-purpose entities to hold the land and physical infrastructure, then secure financing from banks and private investors based on long-term leases.
For lenders, the key is predictable cash flow. If OpenAI, Microsoft, or major cloud providers sign 15- to 20-year leases, such contracts could support billions of dollars in debt. The data center facilities, power rights, and equipment would also be included as collateral.
The article cites that SB Energy, backed by SoftBank, is advancing a project in Ohio, USA, for OpenAI. NVIDIA has agreed to provide up to $105 billion in credit support for the site’s lease and infrastructure, as well as direct investment in SB Energy. The facility is designed to scale up to a maximum capacity of 8 gigawatts of power.
NVIDIA's role is changing
The article argues that one of the most unusual shifts is NVIDIA transitioning from being merely a chip supplier to taking on a dual role as both a supplier and a financing partner. In addition to selling GPUs, NVIDIA is investing in AI labs, infrastructure companies, and cloud service providers, and helping customers secure financing through guarantees and other means.
The article states that this arrangement helps expand hardware demand, as customers find it easier to secure construction funding. However, the market is also beginning to question: if chip suppliers are simultaneously investing in customers, supporting leases, and participating in project financing, how much of the demand is genuinely driven by the market versus being fueled by the suppliers’ financing capabilities?
NVIDIA opposes describing this model as "circular financing." However, the article notes that related concerns have risen to a level significant enough to affect business arrangements, prompting NVIDIA to recently suspend revenue-sharing partnerships with certain small AI cloud service providers.
Constraints may not be limited to the chip alone.
The article also notes that the construction costs of AI data centers continue to rise. In addition to GPUs, memory, networking equipment, transformers, and cooling systems have all experienced supply shortages. Quotes for some NVIDIA-based AI server systems have increased by more than 15% due to rising memory prices.
With interest rates remaining high, financing costs have risen in tandem. The article suggests that if demand for AI continues to grow, Wall Street may view data centers as a new category of infrastructure assets; however, if future AI revenues fall short of expectations, risks will not be confined to technology companies alone but could spread to banks, bondholders, private credit funds, infrastructure investors, and suppliers providing guarantees.