The race to dominate artificial intelligence is reshaping not only the technology industry but also the balance sheets of some of the world’s largest corporations. Behind soaring investment in data centres, advanced chips and cloud infrastructure lies a rapidly expanding layer of financial commitments that remain largely outside conventional measures of corporate debt.
A Nikkei investigation has found that five of the United States’ biggest technology companies — Alphabet, Microsoft, Amazon, Meta and Oracle — have accumulated approximately $1.65 trillion in off-balance-sheet obligations. The figure exceeds their combined reported liabilities of about $1.35 trillion, highlighting the scale of financial commitments tied to the industry’s accelerating investment in AI infrastructure.
The reported obligations have grown sharply over the past four years, increasing roughly eightfold as companies signed long-term agreements to secure graphics processing units (GPUs), lease data centre capacity and finance large-scale computing infrastructure. Under current accounting rules, many of these commitments remain outside corporate balance sheets until facilities become operational or specific contractual conditions are met.
The timing of the findings coincides with the earnings season for the technology sector, with four of the five companies expected to report financial results in the coming weeks. While published balance sheets continue to show levels of debt that appear manageable under existing accounting standards, the investigation suggests that substantial contractual commitments are disclosed primarily through notes accompanying financial statements rather than being recognised as formal liabilities.
The accounting treatment reflects the structure of many AI-related investments. Companies frequently enter into long-term contracts to secure specialised computing equipment or lease facilities years before projects become fully operational. Until accounting thresholds are met, those obligations remain outside the liabilities reported on balance sheets, despite representing significant future financial commitments.
The investigation points to several examples of how these arrangements are structured. Oracle’s expansion of its Stargate project with OpenAI through externally leased facilities allows related obligations to remain outside its reported liabilities under current accounting treatment. Similarly, Meta’s development of a data centre campus through a joint venture in which it holds a 20 per cent stake means associated financing does not appear as debt on the company’s balance sheet.
Among the companies examined, Meta reportedly carries the largest volume of off-balance-sheet commitments. According to the investigation, its hidden obligations total approximately $420 billion, nearly three times the level of debt reported on its balance sheet.
Oracle has recorded one of the fastest increases in such commitments. The investigation estimates that the company’s off-balance-sheet obligations have expanded more than thirtyfold during the past four years to approximately $273 billion, driven largely by its rapid investment in AI infrastructure linked to the Stargate initiative.
The growing reliance on these financing structures has attracted attention beyond the technology sector. The Bank for International Settlements (BIS), often described as the central bank for central banks, examined the practice in a report published in March 2026. BIS economists referred to these arrangements as “shadow borrowing” and warned that they could create new channels through which financial stress spreads across the broader financial system.
According to the BIS report, these obligations are increasingly linked to private credit funds, insurance companies and banks that provide financing for AI infrastructure. As investment in large-scale computing facilities accelerates, financial exposure extends beyond technology companies to institutions that fund or insure those projects.
The scale of global investment reflects the industry’s continuing expansion. Spending on data centres worldwide is projected to reach as much as $7 trillion by 2030, making AI infrastructure one of the largest capital investment cycles in recent decades. That rapid growth has increased demand for financing arrangements capable of supporting projects whose costs often run into tens of billions of dollars.
The investigation notes that while these commitments remain outside reported liabilities under current accounting standards, they nevertheless represent substantial future obligations linked to long-term contracts. As companies continue to expand AI capacity, the volume of such commitments has risen in parallel with growing demand for computing power and specialised semiconductor technology.
The emergence of increasingly large off-balance-sheet obligations illustrates how accounting treatment and financing structures are evolving alongside the AI industry itself. Long-term contractual commitments, once a relatively limited feature of corporate reporting, have become central to financing some of the world’s largest technology infrastructure projects.
With major technology companies preparing to release their latest financial results, attention is expected to remain focused not only on reported earnings and balance sheets but also on the contractual obligations disclosed outside them. The rapid expansion of AI investment has placed these off-balance-sheet commitments at the centre of discussions about corporate financing, financial transparency and the long-term costs of building the infrastructure underpinning the next generation of artificial intelligence.
