$ORCL

The market's bull case rests on Oracle's headline remaining performance obligations (RPO) figure, which reached $664 billion as of the most recent quarter (per the Q1 FY2027 8-K, September 10, 2026). The problem is what lies beneath that number: S&P Global estimates that OpenAI alone accounts for roughly half of the $638 billion backlog that existed at the time of its July 9, 2026 downgrade (per S&P Global Ratings, July 2026). That is not a diversified order book; it is a single-counterparty concentration risk dressed up as a pipeline. The balance sheet deterioration is severe and accelerating. Oracle ended Q2 FY2026 (November 2025) with approximately $108 billion in debt, up from $92.6 billion in May 2025, after completing an $18 billion bond sale in September 2025 (per Motley Fool, December 2026; per Oracle 10-Q, November 2025). S&P now projects total debt obligations of $160-167 billion and a free operating cash flow deficit of negative $42 billion in fiscal 2027, against a prior estimate of negative $24 billion (per S&P Global Ratings, July 2026). Capital expenditures for the first half of FY2026 alone were $20.5 billion, more than triple the $6.3 billion spent in the same period a year earlier (per Oracle 10-Q, November 2025). FY2026 capex guidance stands at $50 billion, double the $25 billion guided as recently as June 2025 (per Seeking Alpha, January 2026), and S&P now projects FY2027 capex of $90-95 billion. The credit market has responded. S&P downgraded Oracle to BBB- on July 9, 2026, one notch above speculative grade, with Moody's and S&P both carrying negative outlooks (per S&P Global Ratings, July 2026; per Barclays, November 2025, which had already flagged the risk of a fall to BBB-). Oracle's debt-to-equity ratio stands at roughly 500%, and its total debt of approximately $100-108 billion exceeds that of Microsoft and Amazon in absolute terms, despite Oracle generating a fraction of their revenue and operating with a far thinner equity cushion (per tomtunguz.com analysis, September 2026). To plug the cash gap, management has already executed a $5 billion mandatory convertible preferred stock offering (February 2026) and has signalled a further $20 billion equity raise, both of which dilute common shareholders (per X/TokenandoAI, July 2026). Management credibility is a secondary but material concern. The company extended the useful life of servers from five to six years in FY2025, reducing depreciation by $181 million per quarter and flattering non-GAAP earnings (per Oracle 10-Q, February 2025). A bondholder lawsuit alleges Oracle misled investors about the scale of planned debt issuance around the OpenAI contract announcement (per X/TokenandoAI, July 2026). R&D spending actually declined 3.6% year-over-year in Q1 FY2027 (per StockTitan financials), suggesting the company is cutting investment in its core software moat to fund GPU-filled data centres. For the short to work, the following must be true: (1) OpenAI remains unable or unwilling to fully fund its own infrastructure needs, keeping Oracle's capex obligations elevated; (2) free cash flow stays deeply negative long enough to force further dilutive equity issuance or a ratings cut to junk; and/or (3) the AI infrastructure cycle softens, leaving Oracle with stranded assets and no ability to service its debt load from operations.