Tech companies have moved more than $120 billion in data centre spending off their balance sheets using special purpose vehicles (SPVs), according to an analysis by the Financial Times, intensifying concerns about the financial risks of the artificial intelligence (AI) boom.
Meta, Elon Musk’s xAI, Oracle, and data centre operator CoreWeave have spearheaded a wave of complex financing deals designed to shield their companies from the enormous borrowing required to build AI-focused computing infrastructure. Financial institutions including Pimco, BlackRock, Apollo, Blue Owl Capital, and major U.S. banks such as JPMorgan have provided debt and equity for these initiatives, enabling tech firms to secure large-scale data centre funding without burdening their balance sheets.
By routing money through SPVs—holding companies specifically set up to finance and own infrastructure—companies can hide the immediate liabilities of their massive AI investments. While this strategy preserves corporate credit ratings and financial metrics, it obscures the underlying risks for both the tech companies and investors, and may pose cascading threats to Wall Street if AI demand falls short.
A senior executive at a major financing institution told the Financial Times that 18 months ago, the scale of such SPV arrangements would have been “unfathomable,” and that today they are “very much the norm.” The executive noted that the tech industry enjoys unrivaled access to private capital due to its strong cash generation and historically low debt, allowing investors to place massive bets on AI infrastructure with confidence.
Meta’s $30 Billion SPV Deal
In October, Meta completed a $30 billion off-balance sheet deal to fund its Hyperion data centre in Louisiana. The arrangement created an SPV called Beignet Investor with New York financing firm Blue Owl Capital. The SPV raised $30 billion, including $27 billion in loans from Pimco, BlackRock, and Apollo, with $3 billion in equity from Blue Owl. Meta retains a 20% stake in the SPV and has offered a residual value guarantee, committing to repay investors if the data centre’s value drops below a threshold at the end of the lease.
The Meta SPV structure enabled the company to borrow $30 billion without the debt appearing on its own books, making it easier to raise a further $30 billion in corporate bonds soon afterward.
Oracle and xAI Follow Suit
Oracle has aggressively pursued off-balance sheet funding for AI-related facilities, particularly for its collaboration with OpenAI. Its deals include $13 billion invested by Blue Owl and JPMorgan for an SPV owning an OpenAI data centre in Abilene, Texas, $38 billion for two sites in Texas and Wisconsin, and $18 billion for a New Mexico facility. In each case, Oracle leases the facilities from the SPVs, meaning lenders have recourse only to the assets themselves, not the operating company.
Elon Musk’s xAI is pursuing a similar structure for a $20 billion fundraise, including $12.5 billion in debt, to purchase Nvidia GPUs and lease them to the company. CoreWeave has also created SPVs for AI contracts with OpenAI, borrowing billions to fulfill obligations and finance computing power for AI models.
Financial Risks and Market Exposure
While these structures allow tech firms to maintain strong balance sheets, they carry hidden financial risks. Investors may assume that the ultimate liability rests with the tech company leasing the facility, particularly if AI demand underperforms, exposing lenders to potential losses. Multiple SPVs across different tech firms and projects could amplify systemic risk, especially in the private credit markets where visibility is limited.
According to UBS, tech companies had borrowed approximately $450 billion from private funds as of early 2025, up $100 billion from the previous year. About $125 billion flowed into long-term “project finance” deals such as the Meta-Beignet transaction, reflecting the rapid expansion of data centre infrastructure fueled by AI demand.
“SPV financings add outstanding liabilities, meaning the overall credit quality for hyperscalers could be worse than currently modeled,” said Matthew Mish, head of public and private credit strategy at UBS. He warned that the combination of high valuations, concentrated borrower exposure, and illiquidity creates a complex risk environment for both investors and companies.
Wall Street and Private Capital Appetite
Financial institutions are eager to participate in the AI infrastructure boom. SPV and project finance deals are attractive to investors seeking exposure to tech giants’ high credit quality while limiting direct corporate liability. Some firms are even exploring securitization of AI debt, pooling loans from multiple SPVs and selling them as asset-backed securities to a broader range of investors, including pension funds and asset managers.
Nevertheless, reliance on SPVs highlights the growing gap between tech companies’ financial statements and the actual risks embedded in their AI investments. While hyperscalers like Google, Microsoft, and Amazon largely finance data centre expansion directly through cash or conventional debt, smaller or newer entrants are increasingly turning to off-balance sheet SPVs to sustain rapid growth and secure computing capacity.
Conclusion
The AI-driven data centre expansion, largely funded through innovative SPV structures, has reshaped corporate finance in Silicon Valley. Companies can now raise unprecedented sums of private capital while shielding their balance sheets, but the hidden risks—both to lenders and to tech companies themselves—are drawing scrutiny from investors and regulators alike.
The Financial Times notes that while the off-balance sheet strategy preserves the financial optics for tech giants, the true exposure to market shifts, AI demand fluctuations, and potential operational setbacks remains substantial. As AI infrastructure spending continues to surge, Wall Street’s involvement in these complex structures could create systemic vulnerabilities, linking the success of the AI boom directly to the stability of global private credit markets.
