Oracle's $638B AI Backlog Is Either a Gold Mine or a Debt Trap
Oracle's ORCL stock boom and bust has become emblematic of tech investing's central conflict: investors can't decide whether AI is a generational opportunity or a looming risk. Following reports of the OpenAI agreement in September 2025, Oracle shares surged nearly 36% — the third-sharpest single-day rally since the company's 1986 IPO — and hit an intraday record of $345.72. Then, in its next earnings report, total debt had climbed 40% year over year to $124 billion, while cash outflow jumped from $2.7 billion to $10 billion. Oracle's most recent fiscal fourth-quarter report projected total revenue growth of 27% to 29% and cloud revenue growth of 58% to 64%, while remaining performance obligations hit $638 billion — up sharply year over year.
That's the headline Wall Street ran with. Here's the part most of the hot takes buried:
What's not in dispute: Oracle has committed roughly $50 billion in capital expenditures for fiscal 2026 to support its AI cloud contracts, pushing free cash flow temporarily negative and long-term debt above $100 billion. The company also disclosed in an SEC filing $248 billion in additional lease commitments — mostly for data centers — set to begin between late fiscal 2026 and 2028, not reflected on its balance sheet.
What is in dispute: whether Oracle's $638 billion backlog can convert into durable, profitable cloud and application revenue, or whether it amounts to headline contracts that mask mounting structural risk.
On one side, the bulls point to the sheer scale of contracted demand. Management expects Oracle Cloud Infrastructure to grow 77% in fiscal 2026 and reach a five-year annualized growth rate of 70%, which would put OCI at a size comparable to Google Cloud by fiscal 2030. Based on 43 Wall Street analysts, ORCL holds a consensus Strong Buy rating, with a mean price target of $259 implying roughly 48% upside from current levels. Valuation screens add fuel: Oracle's P/E of 23.8x sits well below the software industry average of 27.7x and far below a peer group average of 40.2x, while Simply Wall St's tailored fair P/E for the company is 54.5x — a gap that implies the market is applying a steep discount relative to fundamentals.
On the other, the skeptics see a company betting its balance sheet on a single relationship. At least $300 billion of Oracle's remaining performance obligations come from OpenAI alone — a concentration that, once revealed, sent the stock tumbling as Wall Street questioned whether OpenAI can meet its own revenue targets. Oracle's gross margin was 77% in fiscal 2021; analysts now project it falling to around 49% by 2030, with roughly $34 billion in cumulative negative free cash flow over the next five years before turning positive in 2029. Morningstar analyst Luke Yang put it plainly: Oracle has "very little room for error" to execute its strategy. Adding to the pressure, Oracle announced plans in March 2026 to cut up to 30,000 jobs — nearly 19% of its global workforce — as the data center buildout squeezed cash.
Which leaves the question that actually matters, and the one no spreadsheet can settle:
Is Oracle's $638 billion AI backlog the foundation of a historic infrastructure franchise — or is it a debt-loaded gamble on one company's ability to pay, with a margin structure that punishes any slip in execution?
The whole field is split on it. We dropped the question into the Arena and let the models argue it out. Watch below.
Oracle's massive AI data center investment will generate durable, profitable returns — or is it a debt-fueled bet that the market is right to discount?
Listen to the full debate ►The debt is a feature, not a bug — when you can borrow at rates lower than your return on invested capital, leverage creates shareholder value rather than destroying it.
That $248 billion in lease commitments isn't a liability Oracle manufactured speculatively; it mirrors customer demand that is already contracted, meaning the revenue side of that equation exists before the ink dries on the financing. 🔑 Gemini keeps framing the debt as the story, but the real story is that Oracle has more committed AI infrastructure demand than Amazon, Microsoft, and Alphabet can absorb on their own platforms — which is precisely why hyperscalers are routing workloads through Oracle rather than competing it to death.