Oracle

Oracle (NYSE: ORCL) has sold databases since 1977. What put it in this registry is what it borrowed to build cloud capacity for a customer that does not yet earn what it has promised to pay. In fiscal 2026 (ended May 31), Oracle’s capital expenditure hit $55.7 billion, up from $21.2 billion the year before, and free cash flow swung from -$394 million in fiscal 2025 to -$23.7 billion in fiscal 2026, an outflow larger than net income, against +$11.8 billion in fiscal 2024. Borrowings reached $129.5 billion, up from $92.6 billion a year earlier. On 28 October Oracle gave the first date for the outflow to end: the second half of fiscal 2028.

DeadRisk carries Oracle as a second-order, big-tech name inside The AI Capex Supercycle and The Power Wall — not a company whose product is AI, but the clearest example of what a hyperscaler’s balance sheet costs once it becomes one of AI’s biggest landlords. The clearest single thread through that cost runs through Bloom Energy: when Oracle’s gas turbines for a New Mexico data center could not clear permitting, it pivoted to Bloom fuel cells, in a deal signed days after Oracle received a warrant to buy $400 million of Bloom’s own stock. That dossier tells the story from the supplier’s side; this one, the buyer’s.

The origin point has an address: Abilene, Texas. In May 2025, Bloomberg toured the first Stargate site, eight near-identical structures under construction, builder Crusoe spending roughly $12 billion. By July, OpenAI had reportedly agreed to rent an additional 4.5 gigawatts of Oracle capacity, on top of a roughly $30 billion-a-year contract. Two months later, the Wall Street Journal reported a $300 billion, roughly five-year OpenAI contract for Oracle computing power, 4.5 gigawatts starting in 2027; sources said most of Oracle’s newly disclosed revenue would come from this one customer.

What follows can be dated. That it was raised specifically to fund this contract cannot — the record does not establish that link. Within two weeks of the $300 billion report, Oracle sought to borrow $15 billion through a bond sale, as its cash flow turned negative for the first time since 1992. A month later, banks prepared a $38 billion debt sale tied to two Oracle data centers — $23.25 billion for a Texas site, $14.75 billion for one in Wisconsin. By mid-December, a filing showed Oracle had signed roughly $150 billion of new data-center lease commitments in three months, for $248 billion in all. Days later, Financial Times analysis put $66 billion of that construction debt outside Oracle’s balance sheet, in special-purpose vehicles — more than Meta’s $30 billion or xAI’s $20 billion. Oracle’s Q2 revenue came in at $16.06 billion, below the Street’s $16.21 billion, and ORCL fell 11% the next session as Oracle raised its capex forecast to roughly $50 billion, from $35 billion three months earlier.

In January 2026, Oracle took a 15% stake in TikTok’s new US joint venture, alongside Silver Lake and Abu Dhabi’s MGX. The capex story pulled harder. In February, Oracle said it would raise $45 billion to $50 billion in 2026 through debt and equity, and a day later launched a $25 billion bond offering alongside an up-to-$20 billion share program — the largest US high-grade bond deal since Meta’s $30 billion sale the previous October. In March, Q3 revenue beat estimates at $17.19 billion, up 22%, and ORCL rose 9% the next session. Two weeks later Bloomberg reported Oracle and OpenAI had walked away from a planned Abilene lease, which Microsoft took: an expansion beside the first Stargate site, not the site itself.

April brought a Bloom Energy expansion to 2.8 gigawatts. June brought another beat — Q4 revenue of $19.18 billion, up 21% — and another drop: ORCL fell 8.5% on guidance that Oracle would raise nearly $40 billion more in debt and equity in fiscal 2027. Two and a half weeks later came its worst week since 2001, down 19% in five sessions. In July a Nikkei Asia investigation put combined off-balance-sheet debt at Alphabet, Amazon, Meta, Microsoft, and Oracle at roughly $1.65 trillion, up eightfold since 2022; and credit default swap prices tied to Oracle, SpaceX, Alphabet, and Nvidia rose to record highs. Oracle also won a Pentagon agreement worth up to $7 billion over up to ten years.

In the quarter ended 31 August, capex ran $28.5 billion and free cash flow was -$5.4 billion, yet borrowings fell to $125.3 billion from $129.5 billion at 31 May (10-Q filed 11 September). Oracle also finished its $20 billion at-the-market share program. On 28 October, at an investor day, it set the limit (Reuters; Oracle’s release). New capital through fiscal 2030 is capped at $20 billion, $10 billion of equity and $10 billion of gross debt, refinancing excluded. Free cash flow turns positive in the second half of fiscal 2028 and reaches at least $30 billion in fiscal 2029. Capex peaks in fiscal 2027 and 2028. Net debt, leases included, falls below 3.5 times adjusted EBITDA in fiscal 2029. ORCL closed at $144.54 on 28 October against $130.53 the day before, up 10.7% — still 41.8% below its 2026 closing high of $248.15 on 1 June.

On 27 October Oracle added ChatGPT and Codex to its applications, and Google TPUs, an OpenAI chip and Cerebras hardware to Oracle Cloud Infrastructure (CNBC). Two posts on X filed under that story call Oracle the first hyperscaler outside Google to offer its TPUs; one adds a start of calendar 2027. We found neither claim in CNBC, Oracle or Google text, so treat both as posts.

Oracle’s 28 October 8-K (accession 0001193125-26-435853) lists on its cover depositary shares of a 6.50% Series D Mandatory Convertible Preferred Stock. Its only item is the investor day release, which does not mention the preferred. Oracle sold it in February (8-K, accession 0001193125-26-039344): 100 million depositary shares, each a 1/2,000th of a $100,000 share, a $5.0 billion liquidation preference (our arithmetic), carried at $4.954 billion in the 31 August 10-Q. It pays 6.50% a year, in cash, stock or both, through 15 January 2029, then converts into 25.0 million to 31.2 million common shares, set by the average price over the 20 trading days before that date. At an average of $160.06 or less, holders get the maximum.

Situational Awareness LP’s 13F-HR, filed 18 May 2026 for the quarter ended 31 March 2026 (accession 0002045724-26-000008), has one Oracle line: a put referencing $1.07 billion of underlying shares — the value of the stock the option references, not a bet size, not capital at risk, not a short position of that scale. DeadRisk’s analysis found Oracle the largest of three put lines that fell between that period-end and 30 July, down roughly 13% (a close of $147.11 on 2026-03-31 against $127.56 on 2026-07-30, as-traded). The filing cannot say whether it was speculation or a hedge. The next 13F, for the quarter ended 30 June and filed 14 August, has no Oracle line.

TEXXR’s knowledge graph returns 500 hyperedges for Oracle — the API’s maximum, so the true count runs higher. The extraction pipeline classifies 62.0% of them (310 of 500) as financial, 14.4% as partnership, and 12.0% as acquisition. That is a classification frequency, not a measure of Oracle’s debt or capex; it says the pipeline reads most of Oracle’s coverage as financial in kind, consistent with the story above but not proof of it.

The single densest thread in the graph is one address. Eighty-nine edges reference Abilene, Texas: a launch edge from May 2025 naming it the first Stargate site, then a cluster, most dated to one 48-hour window in March 2026 and nearly all marked failed, reporting the same fact in different words: Oracle had abandoned a planned expansion there. That is not 89 events. It is the pipeline registering how hard one reversal hit the record — the inverse of Bloom Energy’s single Oracle edge. A confirmed edge closes the loop: Microsoft agreed to lease the site from Crusoe.

Other edges show Oracle beyond the capex story: confirmed edges for the 15% TikTok stake and the Pentagon’s up-to-$7 billion software agreement, a pair a week apart in late July 2026 for record credit-default-swap prices and the Nikkei off-balance-sheet debt study, and a confirmed launch edge dated 27 October (524561) for Google TPUs.

Positioning: how much of ORCL is sold short, on the record brokers file twice a month. It lags by design — read it as where the crowd stood, not where it stands.

Signals in the coverage and filing record, not trade ideas:

• Whether free cash flow turns on schedule, and what debt does meanwhile — two numbers. Free cash flow ran +$11.8 billion (FY24), -$0.4 billion (FY25), -$23.7 billion (FY26), and -$5.4 billion in the quarter to 31 August. Oracle has now dated the turn: positive in the second half of fiscal 2028, at least $30 billion in fiscal 2029. Borrowings fell from $129.5 billion to $125.3 billion in that quarter; each later 10-Q shows whether they keep falling. The calendar lists the next report on 10 December.
• Whether the $20 billion cap holds. It covers fiscal 2027 to 2030, splits $10 billion equity and $10 billion gross debt, excludes refinancing, and comes on top of the finished $20 billion share program. The release also names vendor financing, customer-provided hardware and prepayments as ways to cut upfront cash, with no figure. Count each new bond or share sale against the $20 billion. The preferred’s January 2029 conversion falls inside the window; the release is silent on it.
• Whether Google TPUs reach Oracle customers in 2027. One post gives that year; an Oracle or Google release would settle it.
• Whether more Stargate sites go the way Abilene did. One planned Abilene expansion already went to another tenant. Other sites — the mid-2025 shortlist included Texas, Michigan, Wisconsin, and Wyoming — are checkable one by one.
• The off-balance-sheet debt trajectory. The Financial Times put $66 billion of Oracle’s construction debt outside its balance sheet in December 2025; Nikkei put five companies’ combined total near $1.65 trillion in July 2026. Whether Oracle’s share widens is a number to track, not infer.
• Whether the market’s reaction pattern breaks. Q2 missed and fell 11%; Q3 beat and rose 9%; Q4 beat and fell 8.5%. An investor day with no earnings print then lifted the stock 10.7%. The next report is the first read against the fiscal 2028 turn date.

Oracle lives or dies by the same constraint as these names — small equal-weighted groups, one constraint each. Not investment products: see all categories.