5 Days Later: Oracle Slides After Blowing Through Entire Four-Year “Hard Cap” In One Morning With $20BN Bond, Stock Offering

Another day, another hyperscaler scrambling to raise capital.

Just one week after Meta tapped shareholders for up to $95 billion, when we said, well, the obvious…

… and just days after Microsoft ended a nearly three-month drought in Big Tech bond issuance with a $25 billion deal, it’s Oracle’s turn. According to Bloomberg, Oracle (ORCL) this morning launched a $10 billion, six-part sale of US investment grade bonds, with maturities from two to 20 years. Initial price talk on the longest tranche is about 3.6 percentage points over Treasuries.

The headlines hit just before 8am:

• *ORACLE CORP FILES MIXED SHELF
• *ORACLE KICKS OFF SIX-PART US DOLLAR HIGH-GRADE BOND OFFERING

And down 1.1% was just the start. Recall that last Wednesday, ORCL exploded higher the moment Oracle said it would limit capital raises to just $20BN for the 2027-2030 period, hitting an intraday high of $148.48. Four trading days later, the stock is at $126.48, down more than 15% from that peak and back below where it traded before the investor day. In other words, the entire “capital discipline” rally has been round-tripped, and then some.

So what’s the problem? The timing. Last Wednesday, at its investor day, Oracle told the market it would limit additional capital raises to $20 billion for all of FY27 through FY30, split $10 billion equity and $10 billion gross debt. The stock soared 10% on the news. Five days later, Oracle is selling… $10 billion in bonds. Put differently, the entire four-year debt allowance is gone before the ink on the investor day slides is dry.

And then, just as we were writing this, the other shoe dropped:

• *ORACLE FILES TO OFFER $10B OF COMMON STOCK

So that’s $10 billion in bonds plus $10 billion in stock, or the full $20 billion “cap” for FY27 through FY30 raised in a single morning. Which, in retrospect, is what that Fitch headline was telling us: “capital raise is at $20B, incl $10B equity.” Remember when we tweeted “now let’s see everyone else’s equity offering” after Meta? We didn’t think the next one would show up a week later, with a bond deal attached. That’s FOUR YEARS OF FUNDING IN FIVE DAYS!

And, at the risk of being pedantic, this is the second Oracle funding promise broken this year. Back in February, when a panicking Oracle unveiled a plan to raise up to $50 billion (Feb 1), the company promised “a single, one-time issuance of investment-grade senior unsecured bonds” in early 2026 and said it did not expect to sell any more bonds this year. It sold $25 billion that month. Today’s deal is bond sale number two… in 2026.

To be fair, the new cap “excludes refinancing of existing debt”, so maybe Oracle will argue that some of the bond proceeds go to refi. Money is fungible, though, and every dollar refinanced with new bonds is a dollar of operating cash that goes into capex instead. And you can’t call a $10 billion stock sale “refinancing” either way.

“De-Risked”… For Five Days

The reversal is especially awkward for the sell-side (which stands to lose billions in advisory fees should the AI bubble pop), which desperately cheered ORCL’s investor day. Goldman’s TMT specialist sales desk wrote last Thursday (available to pro subs) that Oracle had

“de-risked the AI infrastructure build by putting a hard $20B cap on additional equity/debt raises through FY30, guiding to free-cash-flow positive in H2 FY28 and $30B+ FCF in FY29, with growth becoming fully self-funded thereafter… This directly addressed the market’s main overhang of open-ended funding needs and dilution.“

The “hard” cap, it turns out, can be filled up in a single morning. And the “overhang of open-ended funding needs and dilution”? Well, at least it’s no longer open-ended: Oracle got all of it done at once.

Credit got the same sugar high. Goldman IG trader Brad Shelofsky wrote in his weekend IG Credit Week in Review (also available to pro subs) that Meta’s equity plan and Oracle’s “underwhelming $20bn of total capital raise through FY2030″ together “played out some of the bull case, quicker path to self-funding and less bond issuance vs very high expectations.” At Wednesday’s brief tights, Oracle’s 6.7% 2056s traded at T+265bps, 70bps tighter from the prior Friday.

Compare that with today’s ~T+360 price talk on the 20-year tranche. A shorter bond, a wider spread, and a new-issue concession that probably looks a lot like the market pricing in Oracle’s “cap.”

The Self-Funding Mirage

None of this should come as a surprise to anyone who has looked at the cash flow statements. Per Shelofsky, after last week’s earnings Meta moved into negative FCF, “joining GOOGL and AMZN from last quarter, leaving MSFT as the only hyperscaler with positive FCF.” The group as a whole burned $31.4 billion in Q3:

Oracle, for its part, burned $23.7 billion in free cash flow in fiscal 2026 and another $5.4 billion in June-August, and doesn’t expect to turn FCF positive until the second half of FY28 (narrator: read 2030 at the earliest). Meanwhile, hyperscaler capex hit $230 billion in Q3 alone, up 118% y/y, with Oracle’s slice the largest it has ever been:

Shelofsky also notes that Microsoft’s $25 billion deal headlined a $60 billion week in IG primary, nearly double the $30-35 billion estimate, pushing YTD issuance above $2 trillion and past 2020′s full-year record with two months to go. The MSFT deal also took global hyperscaler issuance to $254 billion, above Goldman Research’s $250 billion year-end target. Make that $264 billion as of this morning, and counting.

Everyone Into The Pool

We have been warning since last October that AI had quietly become a debt bubble too, and that whenever one funding window closes, the hyperscalers simply move to the next one. Last week Meta went to shareholders; this week Oracle went to bondholders and shareholders, on the same morning, five days after promising it wouldn’t need much more from either.

Shelofsky’s own near-term view is to “sell strength/rallies” in hyperscaler credit, given “the forward pipeline.” After today, we’d say that pipeline just got a little longer, and the next “hard cap” a lot less believable. Oracle now has zero room left under its FY27-FY30 cap, with capex still rising and FCF negative until the second half of FY28, so we’ll make a prediction: the cap gets “revised” well before FY30. As for Oracle’s promise of self-funded growth by FY29, well, we’ll believe it when the company goes five days without a capital raise.

Much more in the full Goldman “IG Credit Week in Review” and “TMT Spec Sales” notes, both available to pro subs.