Meta, XAI Spread Risks of AI Splurge with Off-Balance-Sheet Debt (www.bloomberg.com)

Bloomberg reports that Meta and X AI have been accelerating AI investments while routing some of the associated costs off their balance sheets, a move that effectively spreads financing risk across suppliers, lessors and other counterparties rather than recording it as direct corporate debt. By using structures such as vendor financing, long-term cloud commitments, specialized leases or third‑party special purpose entities, these firms can expand compute and data‑center capacity for large‑scale model training without immediately showing the full liability on their books. This matters for the AI/ML community and markets because it obscures the true capital intensity and financial risk of scaling foundation models. Off‑balance arrangements can make profitability and leverage metrics look healthier even as real economic obligations grow — raising potential contagion to hardware vendors, cloud providers and financiers if demand or returns for costly training runs disappoint. Technically, the trend underscores how surging demand for GPUs, custom accelerators, power and cooling for training/serving infrastructure is being funded indirectly; researchers and practitioners should expect continued pressure to justify compute efficiency, model compression and cost‑aware ML practices as stakeholders push for transparency and stress testing of AI investments.