The global demand for artificial intelligence (AI) is igniting an unprecedented construction boom for data centers, creating a colossal funding challenge that could reshape capital markets. Projections from leading financial institutions and consulting firms indicate a staggering investment of $5 trillion to $7 trillion globally for data center build-outs between 2026 and 2030. This translates to an annual capital expenditure of approximately $1 trillion, a figure that dwarfs many national economies. For context, Bank of America estimates that constructing one gigawatt (GW) of data center capacity costs $50 billion. At this rate, the projected annual investment would fund roughly 20 GW of new capacity, which is more than three times the total installed electricity capacity of New York City.
This surge in AI-driven infrastructure investment is pushing the boundaries of traditional financing. Hyperscale technology giants, including Meta, Microsoft, Amazon, and Alphabet, are at the forefront, leveraging their substantial free cash flows – estimated to be around $500 billion collectively in 2025 alone – to fund early AI initiatives. As the potential for AI-related breakthroughs became more apparent, the debt capital markets also began to play a significant role. However, as these companies face constraints related to balance sheet leverage ratios, innovative off-balance sheet financing structures are gaining traction.
A prime example of this trend is Meta’s Hyperion Data Center in Louisiana. In October 2025, Meta partnered with Blue Owl Capital to establish a special-purpose vehicle (SPV), Beignet Investor, to develop the project. This SPV successfully raised $30 billion, comprising $27 billion in loans from private credit funds and $3 billion in equity from Blue Owl, effectively moving a substantial portion of this investment off Meta’s core balance sheet.
Looking ahead, financial analysts are anticipating a continued reliance on creative funding mechanisms. Morgan Stanley’s July 2025 report projected that data center capital expenditures could reach approximately $3 trillion by 2028. While hyperscalers are expected to cover half of this amount through their internal cash flows, and corporate debt issuance could contribute another $200 billion, a significant portion – an estimated $150 billion – could be financed through asset-backed securities (ABS) and commercial mortgage-backed securities (CMBS). This points towards a growing importance of structured finance in meeting the immense capital needs of the AI revolution.
Data centers are not entirely new to the securitization market, although their use of structured finance has historically been limited. The pioneering data center ABS issuance occurred in February 2018 when Vantage Data Centers, a prominent operator in key US markets, raised $1.125 billion through rated notes. This capital injection facilitated Vantage’s expansion into existing and new markets. More recently, in 2021, Blackstone issued its first CMBS, securing $3.2 billion to support its $10 billion acquisition of data center operator QTS Realty Trust.
In Europe, the adoption of data center securitization is still in its nascent stages. Vantage Data Centers again led the way in June 2024 by raising £600 million in securitized term notes for two data centers in Wales, marking the first European data center ABS. This was followed by another issuance in June 2025, where Vantage secured €640 million for four data centers in Frankfurt and Berlin, representing the first such deal in continental Europe. The US market has seen more activity, with the New York Times reporting that 27 data center ABS deals were issued in 2025, raising $13.3 billion – a 55 percent year-over-year increase.
The fundamental structure of data center operations lends itself well to securitization. Current data center securitization deals are typically issued by operators who possess long-term contractual cash flows derived from lease agreements with co-location clients or hyperscalers. The proceeds from these issuances are generally utilized to refinance existing debt and expand operational capacity.
The evolving landscape of AI development has also given rise to new service models, such as GPU-as-a-Service (GPUaaS). This model allows AI developers to access crucial computing power on a rental basis, shifting infrastructure spending from upfront capital expenditures to more flexible operating costs. This GPUaaS approach is becoming increasingly attractive for AI training and inference.
Leading providers in this space have recently secured substantial long-term contracts with hyperscalers. For instance, Nebius, a neocloud company based in Amsterdam, has entered into an agreement with Microsoft to provide GPU services valued at up to $19.4 billion through 2031, alongside a $3 billion, five-year deal with Meta. These long-dated contractual obligations generate predictable and stable cash flows, creating a strong foundation for securitizing computing power. This mirrors the established practice of data center operators issuing ABS and CMBS to fund their expansion in a highly competitive market. While high-profile securitization transactions based on these computing power contracts have yet to materialize on a large scale, industry insiders suggest that technology bankers are exploring such ABS deals, though specific details remain scarce.
However, the path to widespread securitization of AI-related assets is not without its challenges. Recent market sentiment has shown signs of strain, particularly in the equity capital markets, as investors reassess the sustainability of the massive AI-driven investments. Oracle’s stock experienced a significant decline of 43 percent from its September peak in late 2025, following its substantial $300 billion computing power deal with OpenAI, which spans five years. This shift in market perception has also been reflected in the derivatives market, with Oracle’s five-year credit default swap (CDS) soaring from approximately 37 basis points in July to 151.3 basis points in November 2025, reaching levels not seen since 2009.
Concerns surrounding Oracle’s aggressive expansion strategy are twofold. Firstly, there is a perceived mismatch between the duration of Oracle’s lease commitments, which are expected to span 15 to 19 years, and the tenor of its contracted revenues, most of which are due within the next five years. This exposes Oracle to renewal risks and potential overcapacity if AI demand softens. Secondly, the company may face challenges with the depreciation of its Graphics Processing Units (GPUs) and the potential need for server upgrades mid-lease. While Oracle currently depreciates IT equipment over six years, similar to its peers, the long-term lifespan of GPUs remains an open question, especially given that cutting-edge AI models like ChatGPT have only been widely available for a few years. Similar pressures have impacted other players in the AI infrastructure sector, with CoreWeave and Nebius shares also experiencing significant declines from their respective peaks by the end of 2025.
Despite these sharp equity corrections, the broader outlook for AI infrastructure funding remains robust. Analysts at Goldman Sachs, for instance, expressed continued confidence in the sustainability of AI funding in December 2025. They highlighted that, in aggregate, 90 percent of AI capital expenditures to date have been financed by the operating cash flows of hyperscalers, with only 10 percent sourced from corporate debt. Notably, a significant portion of this corporate debt has been issued by Meta, a company whose credit ratings exceed those of the US government.
For computing power ABS and other more complex financing vehicles to achieve widespread adoption, clearer evidence of AI monetization will likely be a prerequisite. In the interim, the development of a viable business-to-business-to-consumer (B2B2C) model, where AI adoption demonstrably translates into enhanced productivity and margin expansion for end customers, could serve as a crucial catalyst. Once these underlying economic fundamentals are firmly established, computing power ABS deals could transition from niche instruments to mainstream financing options, diversifying the capital toolkit available to the burgeoning AI industry.
