Is computing power ABS on the horizon to fund AI?

Core Insights - JP Morgan estimates a global expenditure of $5 trillion on data center construction from 2026 to 2030, while McKinsey projects this could reach $7 trillion, indicating an annual capital expenditure of approximately $1 trillion [2] - Bank of America highlights that building one gigawatt (GW) of data center capacity costs around $50 billion, suggesting that $1 trillion in annual investment could fund about 20 GW of new capacity, which is three times New York's current installed capacity of approximately 6.7 GW [3] Investment Dynamics - In 2025, significant investments are being made in AI and data centers, with companies like OpenAI, Oracle, and SoftBank committing $500 billion to the Stargate project over four years, and CoreWeave signing a $14 billion deal with Meta for computing power [4] - The funding landscape for AI investments is complex, with hyperscalers like Meta able to leverage their balance sheets for large commitments, while standalone AI developers face challenges in financing their investments [5][6] Financing Mechanisms - Morgan Stanley projects that by 2028, data center capital expenditures could reach $3 trillion, with half potentially funded by hyperscalers' cash flows and an additional $200 billion through corporate debt issuance [8] - Data centers have begun to utilize off-balance-sheet financing, with the first data center asset-backed securities (ABS) issued in 2018, and a notable increase in ABS deals in 2025, raising $13.3 billion, a 55% increase year-over-year [9][11] Market Trends - The GPU-as-a-Service (GPUaaS) model is emerging, allowing AI developers to access computing power on a rental basis, which shifts spending from upfront capital expenditures to flexible operating costs [13] - Long-term contracts with hyperscalers are being secured by neocloud providers, creating predictable cash flows that could facilitate securitization similar to traditional data center financing [14] Challenges and Concerns - Recent market corrections have raised concerns about the sustainability of AI-driven investments, with Oracle's shares dropping 43% and significant lease commitments leading to potential renewal risks [16][18] - Despite equity corrections, the overall funding landscape for AI infrastructure remains robust, with 90% of capital expenditures funded by hyperscalers' operating cash flows [20] Future Outlook - For computing power ABS and other financing vehicles to gain traction, clearer evidence of AI monetization is needed, with a viable B2B2C model potentially serving as a catalyst for mainstream adoption [21]