Oracle once sold databases. Now it races to erect data centers packed with graphics chips for the hungriest artificial-intelligence projects on the planet. The shift has delivered explosive cloud growth. It has also produced negative free cash flow, repeated bond sales and a credit market that watches the company’s every move like a barometer for AI risk.
Investors felt the strain again this week. Oracle reported fourth-quarter results that beat expectations, with cloud revenue up more than 47 percent and OCI revenue nearly doubling. The company’s remaining performance obligations swelled to $638 billion. Yet shares fell sharply after hours. The reason sat in plain sight. Management signaled plans to raise roughly $40 billion in fresh debt and equity to fund even more capacity. Another massive outlay. Another test of how far markets will stretch for the AI boom.
The numbers tell a story of ambition outrunning near-term cash. Capital expenditures have climbed from single-digit billions to a guided range around $50 billion for the current fiscal year, according to multiple reports. Total debt now exceeds $124 billion after a series of jumbo offerings. Free cash flow turned negative by nearly $25 billion over the trailing 12 months. Cash on hand hovers near $10 billion. The gap is filled with borrowed money.
Gizmodo first highlighted how these escalating commitments rattled traders. The piece captured the market’s immediate upset. But the pattern has only intensified. In September 2025 Oracle sold $18 billion in bonds, one of the largest tech debt deals of the year, Bloomberg reported. Demand reached nearly $88 billion. Early this year the company followed with a $25 billion offering that drew $127 billion in orders, according to the Financial Times.
Oracle’s own February announcement laid out the plan in careful language. The company expects to raise $45 billion to $50 billion in gross proceeds during calendar 2026. Half would come from a single, one-time investment-grade bond issuance. The rest from equity or equity-linked securities. Executives stressed the goal was to maintain a solid investment-grade balance sheet while meeting contracted demand from customers that include OpenAI, NVIDIA, AMD, Meta, xAI and TikTok. “This funding plan reflects Oracle’s commitment to maintaining an investment-grade rating, prudent capital allocation, balance sheet strength, and transparency with investors,” the statement read.
Yet transparency has not quieted concerns. Credit default swaps on Oracle’s debt spiked to levels not seen since the 2008-2009 financial crisis. The cost to insure against default for five years climbed as high as 1.41 percentage points. Banks have struggled to syndicate loans tied to the company’s projects. A $300 billion computing partnership with OpenAI, first disclosed last fall, has become a case study in financing friction. The Wall Street Journal detailed how JPMorgan Chase and others spent months trying to offload risk on billions in loans for data centers in Texas and Wisconsin. Many institutions hit internal limits on exposure to a single name. Balance sheets clogged. Future deals grew harder to finance.
Moody’s and other rating firms have flagged the risks. One analysis cited by Inc. Magazine noted Oracle’s debt-to-equity ratio near 415 percent, far above peers such as Amazon or Microsoft. The company carries more than $160 billion in total liabilities with less than $40 billion in cash. Remaining performance obligations stand enormous, but converting them to recognized revenue takes time. Depreciation on the new infrastructure adds another drag.
But the demand side looks undeniable. Oracle Cloud Infrastructure has emerged as a critical alternative for companies seeking multicloud AI capacity. Backlog growth of more than 360 percent in the latest quarter signals contracts already signed. Revenue from cloud services hit $9.9 billion. OCI alone contributed $5.8 billion. Those figures explain why CEO Larry Ellison and his team keep accelerating spend. They see a window to capture share before competitors finish their own builds.
Still, patience wears thin on earnings calls. Analysts press for clearer timelines on when the heavy capital outlays turn into sustainable free cash flow. Citi’s Tyler Radke has modeled $20 billion to $30 billion in annual debt raises for the next several years. Recent Reuters coverage captured the bond market reaction to reports of another $38 billion debt addition. Yields rose. Prices fell. Investors questioned accounting for depreciation on the AI assets.
Power constraints, local opposition to data centers and the sheer scale of chip purchases compound the challenge. Oracle reportedly agreed to buy $40 billion worth of Nvidia GB200 processors for one OpenAI-related site alone, the Financial Times reported last year. That order sits inside a broader wave of hyperscaler spending that could total trillions over the next five years, according to Fortune’s analysis of Moody’s data. The five big spenders, Alphabet, Amazon, Meta, Microsoft and Oracle, committed nearly $1 trillion combined, much of it still to be spent.
Oracle’s approach stands apart. While Microsoft and Meta fund much of their expansion from operating cash flow, Oracle leans harder on debt and leasing structures. Its $248 billion in undisclosed future lease commitments dwarfs many peers. The strategy has produced results in the income statement. It has also produced volatility in the stock. Shares erased more than $100 billion in market value on one sharp drop last December after spending details emerged. The decline continued into 2026, with the stock down more than 25 percent for the year at points.
Class-action lawsuits have followed. Plaintiffs accuse the company of downplaying risks tied to the spending pace, debt load and timing of returns. The complaints focus on disclosures around capital allocation and credit quality. Oracle has not commented publicly on the suits.
So the questions linger. How long can Oracle sustain negative free cash flow while convincing bond buyers that the AI revenue will eventually materialize at scale? Will banks continue to absorb the loans tied to these projects, or will syndication problems force higher costs or slower growth? And at what point does the market decide the $300 billion OpenAI bet looks less like vision and more like overreach?
Recent trading shows the tension. After Oracle’s latest results, shares dropped more than 7 percent in after-hours action even as cloud metrics impressed. The spending signal overshadowed the growth numbers. On X, traders noted the irony of laying off thousands of workers months earlier to conserve cash while committing tens of billions to concrete and silicon.
Management insists the math works. Contracts carry pricing that should deliver strong returns once utilization climbs. The $638 billion backlog offers visibility few competitors match. OCI’s 93 percent growth rate suggests the infrastructure is finding takers. Yet visibility is not cash. And cash, right now, is leaving the building faster than it arrives.
The coming quarters will test whether Oracle can thread the needle. Raise the promised capital without cratering its credit metrics. Convert enough of that backlog into revenue to bend the free-cash-flow curve higher. Keep hyperscaler partners convinced that Oracle remains a reliable, if expensive, partner in the AI race. Fail on any front and the debt burden could weigh heavier than the growth narrative.
For an industry that has grown used to ever-rising AI forecasts, Oracle has become an unwilling stress indicator. Its balance sheet, its borrowing costs, its stock price all move in lockstep with broader questions about who ultimately pays for the infrastructure that powers tomorrow’s models. The answer, so far, is bond buyers, equity investors and, increasingly, Oracle itself.
WebProNews is an iEntry Publication