New York, New York
— March 9, 2026

Oracle Isn’t Going Bankrupt. The Market’s Pricing It Like It Is.

Oracle has been one of the most volatile mega-cap names over the last year, and the debate keeps circling back to one question: how real is the $300 billion OpenAI deal?

That contract, first reported last fall, is enormous by any standard. It’s also become the single narrative fulcrum for Oracle’s stock. When investors believe it, Oracle trades like an AI infrastructure winner. When doubts creep in, Oracle trades like the balance sheet can’t handle the ambition.

The bear case: credibility risk + debt risk

Skeptics have two main arguments.

First: deal credibility. OpenAI’s competitive landscape is more intense than it was even a year ago (think Gemini & Claude), and the market is hypersensitive to any sign that demand could slip or sourcing could diversify. This week’s reporting that Oracle and OpenAI scrapped plans to expand their flagship Abilene, Texas data center site poured fuel on that skepticism, even as coverage notes the broader 4.5GW capacity agreement remains intact.

Second: the balance sheet. Oracle has levered up to build “AI factories” at scale, which is exactly what you would do if you’re trying to become a serious compute landlord. But leverage makes markets impatient. The more debt you take on, the less tolerance investors have for uncertainty in any single anchor customer.

The CDS freakout: signal or distortion?

The part that really spooked investors recently was credit default swaps (CDS), essentially the “insurance” price on Oracle’s debt. CDS spreads moved higher, and the market treated it like a red flag.

Here’s the key nuance: CDS is not always a clean truth machine, especially in less liquid names or periods of thin trading. That’s where Steve Eisman’s point on Prof G Markets is worth hearing. He argues the CDS market can be extremely illiquid, so it doesn’t take much flow from a single player to move spreads and create a scary chart, even if bankruptcy risk hasn’t fundamentally changed.
Barron’s made a similar point when this started showing up in the narrative, noting CDS levels had jumped sharply and were spooking investors.

Put simply: the CDS spike may be more about market mechanics than Oracle’s actual solvency. It’s a caution flag, not a verdict.

The WSGN thesis: Oracle’s GPU-rental business (GPUaaS) has value with or without one customer

We’ve watched Oracle lose the cloud war to AWS and Azure and know that Larry Ellison is determined not to repeat that story in AI.

Here’s our thesis in one sentence:
Oracle’s ability to rent GPUs at scale is valuable regardless of the credibility of any single customer.

If you believe the world will require AI factories, then ability to deliver compute at scale will be inherently valuable. That “compute landlord” role is durable while customers can come and go like tenants. The OpenAI contract can accelerate it, but it doesn’t define it.

Keep in mind Oracle isn’t building this in a vacuum. The company has been landed smaller significant deals with other firms like xAI and Meta that could easily expand. Any growth in those relationships strengthens the “multi-tenant AI factory” narrative.

Valuation: the gap between fear and reasonable

At current levels (around $150), the market is treating Oracle like a company whose best-case scenario is unlikely and whose balance sheet is a ticking clock.

We think that’s too harsh.

Even Morgan Stanley, which recently cut its price target, still sits at $213. That target implies a world where Oracle executes, the AI factory buildout is monetized, and the market stops treating every headline as existential.

Bottom line

Oracle has real risks: leverage, execution, and the credibility and timing of mega-deals. But “Oracle is going bankrupt” is not the base case, and even prominent credit-focused voices are saying the CDS market can distort the signal.

Our view: Oracle looks oversold, and the market is pricing the story like a single-thread narrative. In reality, AI infrastructure demand is broader than any one customer.

We’ll have more after the earnings call. Be good!