Oracle's AI Infrastructure Spending and Financial Results

Oracle's total debt has climbed to roughly $125 billion, funding a data-center buildout for customers including Nvidia, Meta, OpenAI, AMD and SpaceX ^[1]. That is a rapid climb: the company held about $92.6 billion in debt as of May 2026 and $108 billion by the following quarter after adding $68 billion of new AI-infrastructure backlog with Meta and Nvidia ^[2]. Capital expenditure reached $28.5 billion in the first quarter of fiscal 2027 alone, after fiscal 2026 capex more than doubled year over year to $55.66 billion from $21.2 billion - funded in part by $43 billion in new debt and $5 billion in equity, with roughly $40 billion more borrowing planned for fiscal 2027 and total capex potentially reaching $95 billion ^[3]. Oracle's five-year credit default swaps hit a three-year high, a signal that bond markets are pricing in real funding risk even as the company insists the spending is matched by contracted demand ^[2].

Oracle's headline numbers looked strong: cloud infrastructure revenue rose 121% year over year to $7.4 billion, total revenue reached $19.3 billion (up 30%), and remaining performance obligations - the contracted backlog - hit $664 billion, up $209 billion year over year on more than $30 billion of new AI cloud contracts signed in a single quarter ^[4]. But the company delivered those results while burning cash: free cash flow was negative $5 billion for the quarter, on top of a negative $23.7 billion free cash flow for all of fiscal 2026 ^[3][4]. That leaves Oracle needing to convert a backlog spread across many years into enough real cash to service its debt and lease obligations while capex keeps climbing - the crux of the debate over whether the spending converts into durable earnings or strains the balance sheet before AI demand fully monetizes ^[3].

Oracle disclosed an additional $700 million in restructuring costs in a September 2026 SEC filing, on top of $2.1 billion already accrued, bringing its 2026 Restructuring Plan to roughly $2.8 billion, largely severance payments ^[5][6]. That follows a fiscal 2026 in which global headcount fell by 21,000 even as the company spent $55.7 billion on AI infrastructure, and reports that Oracle managers were drawing up new layoff lists for September 2026 to help fund the AI spending spree ^[7]. The pattern is explicit: restructuring plans are being used to redirect payroll savings toward data-center capex rather than responding to a declining core business.

Oracle's leadership frames the spending as unambiguously justified by demand. Chairman and CTO Larry Ellison has called the AI opportunity's scale 'unimaginable,' while CEO Safra Catz has pointed to signed cloud contracts with 'the who's who of AI, including OpenAI, xAI, Meta, Nvidia, AMD and many others' as evidence the buildout follows real, contracted revenue rather than speculation ^[8]. The clearest expression of that bet is Stargate, the data-center project with OpenAI: a newly announced 4.5 gigawatt partnership adds sites in Texas, New Mexico and an undisclosed Midwest location, bringing Oracle's total Stargate capacity under development to more than 5 gigawatts ^[9][10].

Oracle's expansion is landing inside a broader, increasingly organized backlash against hyperscale AI data centers. Nationwide, 142 protests were held across 42 states in mid-2026, and more than 300 local moratoriums or bans plus over $130 billion in delayed or canceled projects have followed community pushback over water use and land development ^[11]. Industry researchers note that the newest generation of facilities is a different animal from earlier data centers: as one land-use expert put it, 'what we're seeing now is that they're much larger, much more powerful, they're much more resource intensive' ^[11]. That tension is directly relevant to Oracle's own buildout, which includes a new Stargate site in Dona Ana County, New Mexico - the same kind of location where opposition has concentrated elsewhere.