Oracle is not the Bear Stearns of the next financial crisis. But it is an old dog gorging on debt for its next trick

Oracle is not the Bear Stearns of the next financial crisis, but it’s a lot more exposed to a market correction than its cloud-computing peers. And Larry Ellison’s fortunes are now tied to the success—or failure—of a man who might as well be his mirror opposite: Open AI’s Sam Altman.

All the speculation is a function of increasing scrutiny on the financial sustainability of the AI sector. Just this morning tech industry analysts Gartner posted a note warning that the current supply of agentic AI models, platforms and products far exceeds demand, which will result in consolidation and market correction.

According to Gartner its analysts anticipate that agentic AI markets will consolidate in the short term as hype and fear of missing out (FOMO) give way to fundamental economics. “In this AI Vendor Race, the losers of consolidation will be undifferentiated AI companies and their investors. The winners will be capital-rich incumbents with the resources to acquire promising technologies and talent,”

Will Sommer, Senior Director Analyst at Gartner said “While we see early signs of market correction and consolidation, product leaders should recognise this as a regular part of the product life cycle, not a sign of inevitable economic crisis. Over the longer term, consolidation will enable industry leaders to develop agentic products that meet the technical and business requirements of customers who are presently struggling to adopt AI agents.”

It’s the kind of warning that makes investors itchy.

The parallel between Oracle and Bear Stearns is tempting, and right now social media is wallowing in that temptation. Both are, or were, swaggering institutions making audacious bets, each with a balance sheet that invites raised eyebrows. But let’s not get carried away. Bear Stearns collapsed in 2008 because it was a dealer-bank living off overnight money. When that dried up, so did the firm. Oracle, by contrast, makes software, rents compute, and doesn’t rely on repo markets. Still, by any measure, it’s running a riskier balance sheet than its cloud rivals, and its deep entanglement with OpenAI and Nvidia drags it onto a costly treadmill just as markets start to wobble.

✅ FACT CHECKS

1. Oracle’s current and non-current debt totals ~$91bn; equity ~$24–25bn
   ✅ Oracle 10-Q Filing – August 2025
2. Debt-to-equity ~4.5x; peers well below 1x
   ✅ FinanceCharts
   ✅ Macrotrends Microsoft D/E
3. $99.8bn in off-balance-sheet lease obligations (mostly data centers)
   ✅ Oracle 10-Q Filing – August 2025
4. Bear Stearns 2007 leverage ~33.5x; funding via repos
   ✅ Bear Stearns 10-K – 2007
5. Nvidia GB200 NVL72 GPUs deployed on OCI; “thousands” noted
   ✅ Oracle Press Release – June 2025
   ✅ Nvidia Blog – June 2025
6. OpenAI–Oracle $300bn compute deal; 4.5GW capacity
   ✅ WSJ Article
7. OpenAI’s total compute commitments approaching $1tn
   ✅ FT Article
8. Oracle AI GPU rentals: $900m revenue, $125m gross profit (~14%)
   ✅ Seeking Alpha
9. Remaining performance obligations (RPO): $455.3bn as of August 2025
   ✅ Oracle 10-Q Filing – August 2025

Start with the numbers. Oracle’s latest quarterly filing (to August 31 2025) lists $9.1 billion in current debt and $82.2 billion in non-current borrowings—about $91 billion all up—against equity of roughly $24–25 billion. That’s a debt-to-equity ratio somewhere between four and five times. Microsoft, Amazon and Alphabet are all under one. Oracle’s accounts also disclose $99.8 billion in future lease commitments—mostly for data centres—not yet on the balance sheet. That’s not a hidden liability like Bear’s “Level 3 assets,” but it’s still a call on future cash.

Bear Stearns, by comparison, in 2007 had $395 billion of assets on $11.8 billion of equity—a leverage ratio of 33 to 1. Worse, those assets were financial instruments of dubious quality funded with overnight loans. When counterparties stopped rolling repos, the bank simply ran out of money. Oracle’s borrowings are fixed-term notes and leases—no overnight tap to be shut off.

But leverage is leverage. FinanceCharts puts Oracle’s total-debt-to-equity at about 4.5 times as of September 2025, with a 12-month average near 6.9. Macrotrends measures long-term debt alone at 3.3 times. Whichever yardstick you use, it’s a standout. Microsoft, Amazon, Alphabet—0.3 times or less. Oracle goes into any downturn with less equity padding and more interest-rate sensitivity than its peers.

Why? Because it’s sprinting into AI infrastructure at the very moment costs are exploding. In June 2025 Oracle and Nvidia unveiled the arrival of GB200 Blackwell systems on Oracle Cloud Infrastructure (OCI), promising “thousands” of GPUs in Oracle’s data centres. Nvidia’s own blog boasted of liquid-cooled GB200 NVL72 racks on OCI and DGX Cloud. Great technology, expensive habit. GPUs, land, power, and networking don’t come cheap.

Enter Sam Altman—the Anti-Larry. Ellison is the flamboyant hardware baron; Altman is the AI coalition builder who often seems worrying about what he’s created. The Wall Street Journal says OpenAI has agreed to spend $300 billion with Oracle for about 4.5 gigawatts of compute capacity starting 2027—one of the biggest commercial infrastructure deals ever. The Financial Times reports OpenAI’s total compute commitments across partners approach $1 trillion over the next decade. Even a fraction of that would transform Oracle’s top line—and stretch its capital demands.

Nvidia, meanwhile, has announced a letter of intent with OpenAI to supply systems for at least 10 GW of data-centre capacity and to invest up to $100 billion as deployments roll out. That’s vendor financing on a grand scale: good for momentum, but it cements Nvidia’s pricing power. Oracle’s “preferred platform” status may be strategically priceless, but financially it locks the company into an arms race where the invoices arrive long before the profits.

And those profits are thin for now. Investor’s Business Daily, citing internal documents, says Oracle made about $900 million in revenue from Nvidia-powered GPU rentals in its August-ended quarter, but only $125 million in gross profit—a margin of 14 per cent versus the usual 70 per cent elsewhere. Bulls say scale will lift margins; bears say AI IaaS is becoming a low-margin slugfest where cash goes out faster than it comes in. Both might be right. In the short run, Oracle’s AI push will boost revenue and crimp margins at the same time.

Fragile? Not really. Oracle’s debt is mostly fixed-rate and staggered into the 2030s and 2040s. About $9 billion falls due within a year, $82 billion after. Operating cash flow is strong, and its remaining performance obligations—future contracted revenue—hit $455 billion by August 2025. That’s not a house of cards; it’s a company that could see cash flow squeezed if spending runs ahead of earnings.

The real risk is not collapse but compression. Suppose the market falls 10–20 per cent, spreads widen 25–100 basis points, and customers delay projects. Oracle can still refinance, but interest costs rise and the hoped-for AI margins take longer to appear. If the much-talked-about OpenAI deals are delayed, Oracle’s high leverage starts to look less like confidence and more like overreach. On the other hand, if utilisation ramps as planned, the extra EBIT from AI could easily cover the interest bill.

In a full-blown financial shock—a 25 per cent correction, funding jitters, boardroom nerves—Oracle likely remains solvent. Its maturities are long, interest coverage healthy. But deleveraging would stall, credit ratings could darken, and those famous share buybacks might need a rest. The $99.8 billion in lease commitments—“mostly data centres”—could also bite if electricity grid projects lag. Paying to wait for power isn’t insolvency, but it hurts.

None of this makes Ellison’s bet irrational. OCI’s architecture and database strengths give Oracle a real edge in the enterprise AI market. If Altman’s demand materialises and Nvidia’s hardware keeps showing up, Oracle could emerge as one of the big winners of the AI build-out. But the price of that ambition is capital intensity. Big iron needs big balance sheets—and Oracle’s is already bulging.

So, is Oracle the next Bear Stearns? Hardly. Bear was undone by structure; Oracle’s risk is voluntary. One turned collateral into cash until confidence vanished; the other is turning cash into capacity faster than anyone else. That could look brilliant—or very expensive—depending on how the cycle turns. If you’re looking for a historical echo, think not of Wall Street in 2008 but of telecoms in 2001: some built the future and profited; others just built. The difference is execution—and time.

Eleanor Byrne led the reporting for this story.