Oracle's Debt Bomb, Wall Street's Faith, and the Fee You Can Almost Hear Ticking
Oracle's Debt Bomb, Wall Street's Faith, and the Fee You Can Almost Hear Ticking
So let me get this straight. Oracle is sitting on $104 billion in debt. Capex is running $21 billion this year, pushing $35 billion next year. They're borrowing an extra $38 billion on top of the $18 billion they already grabbed in bonds to fund a "Stargate" project with OpenAI and SoftBank. Free cash flow went negative for the first time since 1999—the first time in 26 years. Rating agencies downgraded them to BBB with a negative outlook. Credit default swaps are pricing in what Morgan Stanley calls "crisis-level signals." The cost to insure their debt could hit 200 basis points, a level last seen in 2008 when Bear Stearns was folding.
And yet, there's Deutsche Bank sitting pretty with a $375 target, implying 80%+ upside. HSBC raised their target to $382. Goldman estimates another 18% gain is possible if the AI thesis holds.
This is the most honest debate happening in markets right now. It's not about whether Oracle is good or bad. It's about whether good earnings can exist in a company burning cash while loading up on debt to finance what amounts to a single, massive, utterly undiversified bet on OpenAI's perpetual dominance in AI.
The Honeypot Problem
Here's what Wall Street wants you to focus on: the remaining performance obligations (RPO). It's now $455 billion after tripling in a single quarter. That's $455 billion in contracted revenue that hasn't been recognized yet. It sounds like runway for years.
It is runway for years. If it converts. If it doesn't evaporate.
The problem—and I mean this without exaggeration—is that Oracle is financially dependent on two things happening perfectly. First, OpenAI needs to be spending money at roughly $60 billion a year for the next five years. Second, those other customers need to actually use the capacity Oracle is building for them, not just paper over the contract while they hedge their bets.
OpenAI is burning through $11.5 billion a year and hemorrhaging money. They're not a public company. They don't have the firewall of shareholder pressure to cut stupid capex projects. But they do have limited capital, limited time, and very limited patience for overpaying. If Sam Altman walks into a negotiating room in 2026 or 2027 and says "we've cut our AI spending forecast by 30%," what happens to that $300 billion contract? Does Oracle renegotiate? Do they sue? Do they sit on half-empty data centers?
The market doesn't want to think about this. Wall Street prefers the fantasy where everything goes exponential. Consensus on Oracle is bullish. The stock is up 20% year-to-date despite all this. The fact that it's down 35% from September's peak isn't called a "correction"—it's called an "opportunity."
What the Bond Market Knows
Here's where it gets interesting. The equity market has convinced itself that Oracle is a modern-day industrial giant, justified in borrowing like crazy to build the infrastructure of the future. The bond market is slowly raising its hand and saying: we'd like some actual evidence that you can service this debt.
Five-year CDS on Oracle debt is at 125 basis points. A year ago it was half that. That's the market pricing in real risk. Not the "Amazon in 1999" kind of risk, where growth justifies everything. The "we're not sure you get out of this without financial engineering" kind of risk.
Interest expense is chewing through 20% of Oracle's operating income. It was 10% two years ago. At some point—if capex doesn't generate returns quickly—that math breaks down. You can't service $104 billion in debt on static earnings.
S&P and Moody's both have negative outlooks. That means they're one downgrade away from junk status if anything slips. Free cash flow is in the red for the first time in a quarter-century. As a percentage of operating cash flow, AI capex is consuming 94% of everything the company generates after dividends and buybacks.
Goldman Sachs says the stock could hit $280. Morgan Stanley analyst Brad Zelnick says Oracle will generate $12 billion in free cash flow by 2030 even in a "pessimistic" scenario. That's $12 billion annually on $100+ billion in debt. That's 12% leverage. That's not a pessimistic scenario. That's a default scenario.
Tuesday's Reckoning
Oracle reports December 10. The estimates are consensus: about $1.64 in EPS, revenue in the mid-$16 billions. The stock is trading around $207. If they beat by a nickel, watch the short squeeze carry it higher. If they miss—or worse, if management guides down on capex or talks about "restructuring" their Stargate commitment—bond yields spike and credit analysts start writing pieces about junk-grade trajectories.
The Wednesday Fed cut is basically baked in. Markets are 90%+ confident on that 25-basis-point ease. Which means the only real catalyst this week is Oracle, Adobe, and Broadcom earnings. Broadcom alone is being valued 28% above intrinsic value according to DCF models. Broadcom is down 18% from its highs. Broadcom is also the infrastructure play everyone's hiding behind instead of directly buying Oracle.
So if Oracle's earnings are anything less than "we have this figured out and debt is not a concern," you could see a cascade where the entire AI capex narrative gets revisited. Maybe not immediately. Maybe it takes until January. But the bond market is already voting. The stock market just hasn't caught up yet.
The Real Question
Here's what's actually interesting to watch: whether this is a buying opportunity or a value trap that's learning to smell its own stench.
Oracle is not useless. The database business generates huge margins. The cloud infrastructure deals are real. The customer concentration risk on OpenAI is genuinely mitigated by deals with Meta, xAI, Nvidia, AMD, and others. If this works—if all these data centers fill up, if OCI actually becomes a $144 billion business by 2030—then the leverage was temporary, the margins expand, and you can refinance at lower rates while your free cash flow claws back to positive.
But if Oracle is overestimating demand, overestimating price power, or—let's be honest—overestimating how long OpenAI stays in business or stays solvent, then they've just borrowed their way into a 35% earnings miss in 2027 and a debt restructuring.
Wall Street wants you to pick a lane: bull or bear. But credit markets are sitting in a third lane, arms crossed, waiting to see if you're right before they let you refinance.
On Tuesday, we'll get the first real data point in a month. Not on Oracle's growth. On whether management is starting to feel the weight of that $104 billion anchor. Watch for the tone. Listen to how long they spend defending capex guidance versus defending customer diversification. If Safra Catz sounds relieved, bullish. If she sounds defensive, the bond market gets louder.
The fee is ticking. We just don't know yet if it's a countdown or an hourglass.