Oracle's AI Gamble: Credit Rating Nearing Junk Status, Stock Price Cut in Half
0xBroomberg
Oracle is betting bigger on AI infrastructure than any peer — burning through free cash flow, piling on debt at 4.3× EBITDA, and watching S&P cut its rating to BBB-, one notch above junk, while its stock has nearly halved to $129 since June.
How much is Oracle actually spending?
In the fiscal year ending May 2026, Oracle's free cash flow turned negative. Capital expenditure consumed the bulk of revenue.
The company also signed roughly $260 billion in data-center leases, some kicking in next year.
This means → those leases sit off the balance sheet, making the real burden heavier than the headline debt figure.
FY2027 capex is guided at up to $95 billion; at most $25 billion of that is recoverable from customers.
What are the rating agencies saying?
S&P moved first in July, cutting Oracle to BBB- — one notch above junk.
S&P analyst Andrew Chang projects the debt-to-EBITDA ratio will peak at 4.4× over the next two fiscal years. His line in the sand: "If leverage stays above 4.5×, we may cut again."
Moody's holds Oracle at Baa2 (one notch higher than S&P) but assigned a negative outlook in February, warning leverage "could briefly approach 5×."
Fitch maintains BBB, with a trigger set at sustained EBITDA leverage above 3.5× (or 4× including rent).
How does Oracle compare to its peers?
Oracle carries $129.5 billion in debt at roughly 4.3× EBITDA.
In plain terms = Alphabet, Amazon, Microsoft, and Meta all run the same ratio below 1× — Oracle's leverage is more than four times theirs.
Bond yields tell the same story: Oracle paper trades at 7%–8%, in line with junk-rated debt. Alphabet and Amazon bonds yield just 2.5%–3.5%.
This reflects a market that is already pricing Oracle's debt risk at junk level, regardless of the letter rating.
How real is the "fallen angel" risk?
Morgan Stanley credit analyst Lindsay Tyler flagged the risk plainly: Oracle could carry two low-BBB ratings by year-end, making "fallen angel" concern — a slide from investment grade to junk — reasonable.
Her assessment: "not imminent, but a medium-term risk" hinging on execution and monetization.
This means → if AI data-center revenue fails to materialize before leverage peaks, a further downgrade moves from scenario to probability.
Why are bondholders the most nervous?
Brad Briner, treasurer of North Carolina's public-employee retirement fund, put it bluntly: Ellison "has bet the company before and won, but it's always unnerving for bondholders — we bear the downside and share little of the upside."
In plain terms = if the bet pays off, shareholders win; if it doesn't, bondholders take the first loss.
Oracle told its earnings call it is "committed to disciplined capital allocation and maintaining an investment-grade rating" but did not address the downgrade directly.
Whether the rating holds comes down to a race against time: can AI cash flows arrive before leverage peaks?
Content is for reference only, not financial advice.