Could Oracle Go Bankrupt – Really?
Is it truly possible that artificial intelligence, the very technological wave promised to unlock trillions in enterprise productivity, could end up upending Oracle (ORCL), the staid, unshakeable, fifty-year-old titan of relational enterprise databases? The company that has quietly powered global banking, enterprise logistics, and government administration for half a century is now carrying a very different kind of balance sheet, one built on server racks, liquid cooling pipes, and high-interest bonds.
On paper, the question sounds absurd. Oracle spent decades building one of the most profitable, defensible moats in tech history. Once a Fortune 500 company embeds its mission-critical accounting, ERP, or database systems into Oracle’s architecture, ripping it out is hard. That legacy software business throws off high double-digit margins and billions in reliable cash every year. But that has rapidly changed. Chasing a spot as a dominant hyperscale cloud provider, Oracle has turned itself into a capital-hungry industrial infrastructure company wearing a software company’s stock multiple. See Oracle’s segment financials.
Capex Outrunning Cash Flow
Start with the free cash flow picture, because that’s where the story gets uncomfortable. Oracle is funding much of its AI data center expansion with borrowed money rather than cash generated by the business. Building AI infrastructure is expensive: tens of thousands of GPUs, massive power contracts, and new data center campuses all require huge upfront investment.
The numbers from fiscal 2026 (ended May 31) show just how aggressive the spending has become. Oracle spent $55.7 billion on capital expenditures, up 162% year over year. At the same time, operating cash flow rose a healthy 54% to $32.0 billion. That’s strong growth, but it still wasn’t enough to cover the investment. See how financials of major IT players stack up .
After subtracting capital spending, Oracle generated negative free cash flow of $23.7 billion. Oracle filled much of that gap by raising $43 billion in new debt during the year, taking total debt to roughly $167.4 billion.
Credit markets are beginning to notice. Oracle’s credit default swaps (CDS), which investors use to insure against a company’s default, have climbed to their highest levels on record, exceeding even the peaks seen during the 2008 financial crisis. [1] Rising CDS spreads don’t mean bankruptcy is imminent, but they do indicate that lenders are demanding greater compensation to bear Oracle’s credit risk.
To be fair, Oracle isn’t alone. Hyperscalers such as Google are also spending heavily on AI infrastructure. The difference is that Google generates far more cash from its core business. Alphabet produced $185.7 billion in operating cash flow over the trailing twelve months, giving it much greater flexibility to fund investment without relying as heavily on debt.
The Real Risk Isn’t Debt. It’s Who’s On The Other End Of It
A big chunk of Oracle Cloud Infrastructure’s GPU capacity is leased out to frontier AI labs like xAI and OpenAI, plus a wave of high-growth generative AI startups. Oracle’s fortunes are now tied directly to whether those tenants can actually pay their bills over the life of the contract. Here’s the chain reaction worth watching. It starts with enterprise buyers getting cold feet.
Corporate spending on AI has surged over the past two years, but many companies are still struggling to deploy AI at scale or demonstrate a clear return on investment. At the same time, the U.S. labor market has remained relatively resilient, reducing the urgency to replace workers with AI in many industries. Some large AI adopters, including Tesla, have also begun signaling a more measured approach to AI spending for employees.
If AI tools keep struggling to show clear ROI, corporate software budgets tighten. That cools revenue growth for the AI labs themselves, right as their compute burn rate stays sky-high. From there, the multi-year, multi-billion-dollar compute commitments those labs signed with OCI start to wobble, getting renegotiated or defaulted on outright. And Oracle ends up holding the bag: empty data centers, depreciating chips, and multi-year lease obligations.
Amazon and Google would also feel the impact if AI spending slowed. But their far larger cash flows and more diversified businesses make them much better positioned to weather a downturn than Oracle, whose AI infrastructure business has become a much bigger driver of growth.
Whether AI spending ultimately delivers the productivity gains needed to justify today’s infrastructure boom remains an open question. We discuss that in more detail in The AI Agent Shake-Up: Which Software Stocks Survive?
Why This Actually Matters For Solvency
GPUs age fast. Next-generation chip architectures can make today’s clusters look outdated within three to five years. If anchor tenants default or scale back hard, Oracle can’t just hand the hardware back or cancel those long leases. The commitments are fixed even if the revenue backing them isn’t. Push that further: credit downgrades, tighter credit markets, and refinancing get a lot more expensive, feeding a spiral that’s hard to climb out of. Oracle’s legacy database business remains a valuable cushion, but the AI buildout has introduced a level of financial leverage and capital intensity the company has never had to manage before.
Bankruptcy isn’t the base case. But for the first time in Oracle’s history, the more important debate isn’t about earnings or valuation – it’s about credit. Equity investors can tolerate years of weak free cash flow if they believe growth will eventually justify it. Credit investors cannot.
Investors concerned about concentrated AI infrastructure risk may prefer a more diversified approach. The Trefis High Quality (HQ) Portfolio invests in a diverse group of 30 stocks that have collectively delivered stronger upside with reduced volatility compared to the broader indices. Discover the methodology behind these smoother, higher returns by checking the HQ Portfolio performance data.
Notes: