In November, Meta Platforms Inc. raised approximately $60 billion to build data centers, aiming to secure a leading position in artificial intelligence. Half of this amount will not appear as debt on the social media giant’s balance sheet. Meta is among the companies pushing to keep debt entirely off-balance-sheet, a strategy that allows firms to raise massive capital while minimizing the impact on their financial health. Morgan Stanley structured a $30 billion deal for Meta, the largest private capital transaction in history. This debt will be held in a special purpose vehicle (SPV) linked to Blue Owl Capital Inc. The arrangement also enabled Meta to raise an additional $30 billion in the corporate bond market shortly afterward. Bankers note that off-balance-sheet debt through SPVs or joint ventures tied to assets like chips and real estate is becoming the preferred method for financing AI data center projects. Morgan Stanley estimates that by 2028, tech companies and related parties will need to raise up to $800 billion in private credit through asset-specific deals, including SPV structures. As major tech firms release quarterly earnings, some of their liabilities may be hidden in indirectly controlled transaction structures. AI-related debt is accumulating at a rate of roughly $100 billion per quarter, a pace that "would make anyone familiar with credit cycles raise an eyebrow," said UBS strategist Matthew Mish in an interview. Moreover, the speed of capital raising is accelerating. Off-balance-sheet debt and independent financing vehicles have a notorious history, linked to several high-profile scandals that caught investors off guard. In 2001, Enron Corp.’s off-balance-sheet entities triggered the energy company’s collapse. Years later, banks widely shifted mortgages and other debt into off-balance-sheet vehicles, ultimately triggering a crisis when they were forced to reintegrate these liabilities. Although accounting and rating standards have evolved, the resurgence of financial engineering has led some analysts to question whether all hidden debt can be easily identified. "We’re still in the very early stages of capital raising," said Anish Shah, Morgan Stanley’s global head of debt capital markets. The broader AI ecosystem "will require about $1.5 trillion in external financing, with issuers tapping multiple funding sources." This may force creative solutions regarding debt ownership. For example, Elon Musk’s xAI Corp. is seeking a $20 billion SPV financing deal for its data centers, with the company only responsible for leasing NVIDIA Corp. chips. Traditional Debt Most companies, especially those considered blue-chip or low-risk, can borrow directly in the corporate bond market. Oracle Corp. took this approach, issuing $18 billion in publicly traded bonds in a single day this September to fund its cloud infrastructure expansion. Meta completed the year’s largest investment-grade bond issuance on October 30, just a day after announcing plans to increase spending in 2026. However, such debt financing can erode future borrowing capacity, as companies take on excessive risk. If Oracle were to borrow billions more relative to earnings, it could face a downgrade from investment-grade to junk status, forcing creditors to demand higher interest rates. Companies in the AI race also avoid long-term corporate bonds to prevent being burdened if assets lose value in a few years. "These tech giants can’t predict the AI landscape five years from now," said S&P Global Ratings analyst Naveen Sarma. "That’s partly why they aren’t just issuing corporate bonds—they need flexibility in case data centers become obsolete." Data center operators can also issue structured debt, such as bonds tied to assets like consumer loans or mortgages. For instance, Switch Inc. borrowed millions by packaging data center-related receivables into asset-backed securities, an efficient way to monetize existing financial assets. Yet such debt often remains on balance sheets: Developers sell asset-linked leases, loans, or receivables to financial entities, which then issue bonds. Off-Balance-Sheet Financing Enter Wall Street’s latest solution: Companies like Meta and xAI can raise capital by offloading most risks to third-party investors, minimizing balance sheet and rating impacts. This strategy mirrors energy firms’ long-standing financing for pipelines and renewable projects—third-party capital creates SPVs or joint ventures, forming new legal entities that hold assets like chips or data centers while attracting equity from asset managers or venture capital. These entities issue bonds, often investment-grade due to ties to fast-growing AI firms, raising capital for tech projects. AI companies pay rent or fees to limit financial exposure while benefiting as exclusive tenants. "The scale and creditworthiness of hyperscalers elevate these deals to a different level," said Morgan Stanley’s Shah, citing Meta’s $2 trillion market cap. "This enables far larger financing than before." For xAI, the Musk-backed startup turned to an SPV to secure funding for NVIDIA chips after hitting secured debt limits. Valor Equity Partners and Apollo Global Management led the $20 billion financing via an entity independent of xAI, which will lease the chips exclusively. Valor and NVIDIA provided equity, while Apollo and others lent via investment-grade debt issued by the SPV. xAI’s only risk? A five-year lease—nothing more. Alphabet Inc. took another route: Google guaranteed data center debt for crypto miners, booking it as credit derivatives. If these firms default, the tech giant bears the obligation. Lenders Pounce Private credit firms have waited years for this moment. They’ve raised massive funds from insurers and pensions eager for high-yield, investment-grade debt but previously lacking suitable projects. More capital once flowed to equities, especially tech. NVIDIA’s market cap, for instance, is 45% higher than a year ago. Historically, such growth funded expansion. But AI infrastructure demand now outpaces revenue and stock gains. Companies are locked in an arms race for AI’s next breakthrough—and its profits. Risks remain: Leases may terminate early, investor protections may lag, or assets could depreciate faster. Cloud providers estimate chip lifespans at five to six years, but some become obsolete in three; data centers may be outdated in five. Concerns over debt-fueled growth are spreading, with the Bank of England among those warning. Many recall the dot-com bubble, questioning if the U.S. will repeat history. "During the dot-com era, growth was equity-financed, limiting economic fallout when the bubble burst," said UBS’s Mish. "Now, AI capex is debt-driven—and partly off-balance-sheet." Debt isn’t the only path. Big Tech reinvests hefty profits, while venture capital, sovereign wealth funds, and private equity will inject about $350 billion, mainly into closely held firms. Anthropic’s $13 billion equity raise via Iconiq Capital and Qatar’s sovereign fund falls into this category. But over $1.15 trillion will enter as debt.