Ellison pledges a second issuer’s stock, and the collateral is growing faster than the pledge Oracle Corporation (NYSE: ORCL) · co-founder, chief technology officer & chairman
- What he did
- In the days before his son David’s $110bn Warner Bros. Discovery takeover closed on 6 October 2026, he pledged Tesla, SpaceX and Paramount Skydance shares as loan collateral — the first appearance of any of the three on the public record. Oracle: 413,000,000 shares pledged as of the 21 September 2026 record date, up from about 346,000,000 a year earlier, out of roughly 1.16bn beneficially owned. California lender filings record bank-held security interests in about 87,000,000 Oracle shares, up from about 30,000,000 a year earlier — Citigroup, HSBC, JPMorgan Chase, UBS and BNP Paribas. The family contributed about $17bn of the deal’s $47bn equity at $12 a share, and he gave a personal guarantee covering more than $40bn of the equity financing. On 12 September he cancelled a 10b5-1 plan adopted 22 June 2026 for up to 50,000,000 Oracle shares, one day after a filing disclosed it, with none sold.
- The document
- Oracle DEF 14A, EDGAR-accepted 25 September 2026, 4:05pm ET: the 413m figure, “pledged as collateral to secure personal indebtedness,” with the Governance Committee concluding the arrangements “do not pose a material risk,” that none are in margin accounts, that they are not used to hedge, and that he “has the financial capacity to repay the related personal term loans” without selling. California lender filings for the 87m figure. The Tesla, SpaceX and Paramount Skydance legs rest on press reporting of 20–21 October and carry no share counts in any source I could reach.
Consensus read
Textbook buy-borrow-die. Selling crystallises capital gains and dilutes control, so he borrows instead. Oracle’s proxy says these are plain personal term loans, no margin accounts, no hedging, and the board found no material risk.
Steelman: he does control more than 40% of Oracle, he did keep the shares, and the mechanics are exactly what every adviser recommends to a founder with a concentrated position and a liquidity need. Reaching for a second collateral pool when the first is already 36% encumbered is prudent diversification of lender exposure, not distress.
Contrarian read
This is accommodation, not optimisation. The thing that moved is not the pledge total but the ratio of perfected bank collateral to it: total pledged Oracle grew 19.4% year on year while bank-recorded security interests grew roughly 190%.
That divergence is the signature of lenders tightening advance rates against a borrower whose principal collateral has lost about 62% of its value since September 2025 — more collateral per dollar borrowed. Not a man calmly arbitraging the tax code, but one meeting calls while carrying a $40bn+ personal guarantee on somebody else’s acquisition.
Evidence that discriminates
- The two growth rates diverge. +19.4% in total pledged against roughly +190% in bank-perfected security interests, over the same twelve months. Under the optimisation read both should scale with borrowing, because you post collateral in proportion to the loan. They do not. Either the loan book grew far faster than the pledge disclosure suggests, or the haircut widened. Both readings are bad for the consensus; only one is bad for mine.
- He built an exit and then destroyed it. A plan adopted 22 June to sell up to 50,000,000 shares, cancelled 12 September, one day after disclosure, with nothing sold. Under the consensus read he never wanted to sell — but then the plan should not exist. Optimisers do not file 10b5-1 plans they never intend to use. The plan’s existence is the discriminating fact; its cancellation the day after it became public is the tell.
- New collateral classes rather than more of the old. If the binding constraint were tax, the cheapest move is to pledge more Oracle: he owns about 1.16bn shares and has encumbered only 36%. Instead he reached for Tesla, for a private SpaceX position traceable to a $1bn cheque in the 2022 Twitter takeover, and for a newly listed, thinly traded Paramount Skydance. Reaching past your deepest and cheapest collateral is what you do when the lender will not take more of it.
Honest caveat on the measurement: the two figures come from different document types — a proxy disclosure of total pledged shares, and California lender filings recording perfected security interests. They are not strictly like-for-like, and a rise in perfection can reflect a change in documentation practice rather than a change in terms. The direction is solid; the 190% ratio is indicative, not exact, and I would not build a position on it alone.
What would falsify me
Disclosure that the loan principal grew at roughly the same 190% pace as the perfected collateral — that would make the collateral growth proportionate, and the tightening story collapses.
Lender terms showing unchanged advance rates over the twelve months. Share counts for the Tesla, SpaceX and Paramount Skydance pledges that turn out to be token rather than material. Or Oracle recovering and the pledged count falling back without any sale, which would show the pledge tracks opportunity rather than need.
Pattern — against his own history Pledging Oracle is the decade-long habit and is fully documented: 346m → 413m, 35.7% of the stake. What is new in one week is a second public issuer’s stock as collateral, a private company’s stock as collateral, a personal guarantee north of $40bn, and a cancelled sale plan with nothing sold. Four firsts is not a habit; it is a change of regime.