Last autumn, I wrote a series of notes on credit, beginning with broken retail CLO listed ETFs and Business Development Companies in a Snapcount series (which you can read in order here, here, and here). In December, I followed up with a note on Oracle which I described as an emerging credit risk with contagion potential. Not unlike what happens to ECM managers as the IPO channel becomes stuffed, I noted that there was a growing risk in 2026 where high yield managers might find themselves flipping from buying high yield products at all-time tight spreads (because all the crap moved from junk to private credit over the past five years after Covid defaults largely wiped the HY slate clean), to being forced to make room and sell paper.
Today's note serves as a brief update to this topic as the credit market is beginning to signal that the problems relating to Oracle are starting to tip over, and there are implications for broader risk assets and positioning we need to consider.
Comparing ORCL's financial position since last discussed here, we can see that ORCL's debt (excluding loans and preferred shares) has grown from $100bn in December to $120bn, while the weighted coupon on its debt has moved from 4.4% to 4.8%:
Since my December note, the three key ratings agencies (which all had ORCL at BBB equivalent) are still sitting in the same rating with exception of S&P, which has since downgraded ORCL to BBB- … one level above junk:
All of ORCL's solvency and liquidity metrics have deteriorated since then along with its market cap which at the time was still trading north of $500bln vs $365bn today.
Now keep in mind, back in December the BB spread was trading at around 175bps; today it is closer to 150:
Meanwhile the BB CDX index series from then has largely stayed flat:
So while the HY BB contingent has continued to tighten in spread, what has happened to ORCL's credit in the eyes of the market?
It's not good. Oracle's share price has collapsed below the March lows, while its 5Y CDS has blown out through those March levels to trade at nearly twice the BB CDX index:
While all the major hyperscalers' bonds have been widening and CDS rising since the beginning of the year as they issued massive amounts of debt to fund chips and datacenters, ORCL has required its own axis (remember ORCL is supposed to be investment grade!):
You can see the problem — ORCL's credit is now not only ~200bps wide vs the investment grade complex around ~117bps, but it's wider than the BB corporate complex at 154bps!! ORCL 5Y note in blue, IG in red, BB in yellow, HY in white:
In case you are not fully versed in how deeply the AI debt bubble problem runs at Oracle, I would recommend a read of Ed Zitron's recent note. In the section on Oracle and Larry's Ellison's Fortune, Ed reminds us that Oracle intends to spend at least $90bln in FY2027 (ORCL year-end is May), while ORCL is the hyperscaler most levered to OpenAI given its $300bn+ compute/Stargate commitments.
With each day that passes, it is becoming harder for ratings agencies to justify an investment grade rating. Heck, it would be difficult to justify a BB junk rating here. The market as usual is running well ahead of the agencies.
ORCL's debt pile has grown 20% to $120bn, making it now ~7% of the entire high yield market if it were to be downgraded. And just like When IPOs Forewarn Market Rollovers where supply overwhelms the ability of ECM managers to carry issuance and closes the ECM cycle, high yield managers would have to make a lot of room to catch ORCL paper from a point where HY spreads are trading near all-time tights.
There is another problem with an ORCL downgrade to junk that is less obvious. The typical duration of the entire US HY market is 2.9 years. In other words, existing HY debt is not particularly sensitive to interest rate risk, even if it carries higher credit or default risk.
The average duration of ORCL's publicly traded debt is 7.6 years. This is a mismatch with several problematic implications. For one, it makes ORCL's debt far more sensitive to both a rise in risk-free rates, but also in a deterioration in spread. Furthermore, if HY bond managers are suddenly going to buy ORCL debt upon a downgrade to junk, they will find they have a duration mismatch between ORCL debt and their existing portfolios, requiring a much steeper discount on ORCL debt and creating a potentially significant air pocket (think of what happened to the UK 2061 Gilts or the sovereign century bonds during the 2022-23 bond bear market). The prices on ORCL debt would get absolutely annihilated right when HY bond funds are trying to raise cash and make room among existing holdings. Spreads would widen across the HY complex at a time when dealer balance sheets are pretty heavily extended from leverage to hedge funds running high gross exposures, levered ETFs, and bond basis traders (and Wall Street banks have curtailed their market-making activities in HY cash bonds since Dodd-Frank) — a subject we have discussed before with regards to equity funding.
An Oracle downgrade to junk in 2026 remains a key risk for pushing the already deteriorating credit cycle into another leg down, and I think the credit markets are screaming at us again to be careful on this. Worth thinking through other contagion implications, such as the classic Who's Next notion in the context of Snapcount, such as what happens if people look at ORCL, and wonder if another hyperscaler might be vulnerable, or if widening BB spreads touch off a refinancing issue for significant issuers, or…who knows.
Stay frosty, and as always, kindly yours,
Paulo aka Cloudbear