S&P Global Warns Cloud Giants' AI Data Center Debt May Exceed Disclosures; Oracle Rating Cut to BBB-

S&P Global warned on October 19 that major cloud operators' actual financial obligations tied to AI data center buildouts may exceed currently disclosed levels, as lease, purchase, and financing arrangements obscure true exposure. Mark Habib, S&P's global head of digital infrastructure, said some potential liabilities are not fully disclosed, leaving markets unable to gauge the real risk and the size of the added obligations unquantified. Oracle (ORCL) illustrates the concern. In July, S&P incorporated $260 billion in new lease commitments effective fiscal 2027-2029 and about $13 billion in unconditional purchase obligations as of the end of May, mainly data center power arrangements, into its debt adjustments. In July, S&P also cut Oracle's long-term credit rating to BBB- from BBB, one notch above junk, with a stable outlook. The risk stems from heavy upfront spending on facilities, equipment, and long-term leases while revenue accrues over years. S&P has not concluded that cloud operators' credit risk has broadly deteriorated; the key remains whether AI investment generates sufficient returns.