The financing of the artificial intelligence boom is moving into a new phase, with Wall Street increasingly treating data centers and the computing capacity inside them as infrastructure that can be financed with long-term debt rather than assets that technology companies must fund largely from their own balance sheets.
A recent Securities and Exchange Commission staff position could make that transition easier by giving investors greater flexibility in structuring certain data center financings without triggering the risk-retention requirements that apply to some asset-backed securities.
The move follows a major deal in the AI industry. Nvidia last week announced partnerships with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR to establish financing platforms capable of mobilizing more than $500 billion in third-party capital for AI computing infrastructure. Nvidia said it could provide backstops of up to $125 billion, or roughly a quarter of the potential financing.
Register for the next Tekedia Mini-MBA.
Register for Tekedia AI in Business Masterclass.
Join Tekedia Capital Syndicate and co-invest in great global startups.
The initiative is seen as another indication that the financing of AI infrastructure is becoming almost as important as the technology itself. AI developers and cloud providers need enormous amounts of capital to build data centers, secure electricity and purchase GPUs, but not every customer can finance that expansion directly from its balance sheet.
That is creating an opening for private equity, private credit and structured-finance investors.
“Folks contemplating this transaction will be quite happy about the response from the SEC,” Orion Mountainspring, a securitization attorney at Orrick, told CNBC.
The SEC’s position followed a request from law firm Latham Watkins concerning whether certain data center debt should be treated as an asset-backed security under the Exchange Act. The agency agreed with the firm’s interpretation that the structures under discussion would not fall within that definition and therefore would not be subject to the associated risk-retention requirements.
That matters because Dodd-Frank rules introduced after the 2008 financial crisis generally require sponsors of covered securitizations to retain part of the credit risk attached to the assets they package and sell. Removing that requirement for qualifying data center structures can reduce the amount of equity sponsors need to commit.
Mountainspring said the decision could allow sponsors to “push down the required equity in the deal,” making the structures more attractive to capital providers.
B.K. Lee, an asset-backed securities attorney at Alston & Bird, said the guidance could make data center financing more “flexible and capital-efficient” by reducing the constraints associated with conventional risk-retention structures.
The development also points to a broader change in how AI infrastructure is being financed. Data centers were once largely viewed as specialized real estate and technology assets. Increasingly, investors are evaluating them as long-duration infrastructure capable of producing recurring cash flows from customers that need computing capacity for years.
That shift is already visible in the emergence of dedicated investment vehicles. KKR, for example, launched Helix Digital Infrastructure in June with more than $10 billion of committed long-duration capital to finance data centers, power and connectivity for AI workloads. Nvidia is a founding investor and strategic partner.
The SEC position could add another financing tool to that expanding capital stack.
Why The Financing Shift Matters
The economics of AI infrastructure require enormous upfront spending. A data center developer must often commit capital years before the facility reaches full utilization, while AI hardware also has a comparatively rapid technological replacement cycle.
Structured financing can move some of that burden away from the companies building and operating the facilities. Instead of relying entirely on corporate borrowing or equity, developers can potentially raise money against expected future revenues generated by computing capacity. That could allow more projects to be built simultaneously and give AI companies access to infrastructure without carrying the entire cost on their own balance sheets.
The scale of the capital requirement helps explain why Nvidia is increasingly involved in financing as well as supplying the chips used in AI systems. The company has described computing capacity as an emerging asset class and is working with major financial institutions to make long-term capital available to its customers.
But the model also introduces a new layer of financial risk.
The fundamental question for lenders is whether the future cash flows from AI data centers will be large and durable enough to support the debt being raised today. That depends on continued demand for AI services, high utilization of computing equipment, electricity costs, and the pace at which new generations of chips make older infrastructure less competitive.
There is also a potential circular-financing concern. Nvidia has a direct commercial interest in the expansion of AI infrastructure because more data centers generally mean more demand for its chips. If Nvidia helps support financing for infrastructure that subsequently purchases Nvidia equipment, the company could benefit both as a technology supplier and as a facilitator of the capital used to buy that technology.
That does not by itself make the financing unsound, but it makes the quality of underlying demand particularly important.
Recent developments show why investors are paying attention. Nvidia is providing substantial financial support for an OpenAI data center project in Ohio, including a guarantee of up to $105 billion for the first phase and a $1.5 billion investment in SB Energy. The facility is expected eventually to reach 8 gigawatts of capacity, with the initial phase relying exclusively on Nvidia chips.
The size of these commitments illustrates the financing challenge facing the AI industry. Companies are making infrastructure commitments measured in tens or hundreds of billions of dollars, while investors are being asked to assess demand for computing capacity years into the future.
SEC Guidance Is Not A New Rule
The regulatory change should not be overstated. The SEC position is a staff opinion rather than a formal rulemaking or congressional legislation. It does not create a blanket exemption for every data center financing transaction.
Instead, it gives market participants a clearer indication of how the agency views a particular legal structure.
That clarity can still have a significant commercial effect. Financial institutions and their advisers can now design transactions with greater confidence that structures following the framework outlined in the SEC’s exchange with Latham will not automatically be subjected to the risk-retention regime applicable to Exchange Act asset-backed securities.
Attorneys at Katten Muchin Rosenman said the distinction rests partly on the nature of data center assets and their revenues. Unlike mortgages, which are designed to repay through the gradual liquidation of an underlying asset, data centers can generate continuing operating revenues over their useful lives.
That difference could give financiers greater room to structure debt around long-term computing contracts and other predictable revenue streams. The likely result is a wider range of financing options for the AI infrastructure industry, from conventional project finance and private credit to asset-backed structures and dedicated infrastructure funds.