Oracle’s AI Buildout Tests the Limits of Its Legacy Moat
Oracle’s decision to pour tens of billions into AI data centers has transformed the company from a high-margin software stalwart into one of the most capital-intensive infrastructure players in technology. The shift has produced a stark financial picture: fiscal 2026 capital expenditures reached $55.7 billion, up 162 percent year over year, while operating cash flow of $32 billion left the company with negative free cash flow of $23.7 billion. Capex Outrunning Cash Flow
That gap was financed largely through $43 billion in new debt, pushing total borrowings to roughly $167.4 billion. Credit default swap spreads have climbed to levels not seen even during the 2008 crisis, signaling that fixed-income investors are pricing in greater risk. The stock, meanwhile, has fallen nearly 49 percent over the past year, reflecting market concern that Oracle’s aggressive spending could outrun revenue recognition for years.
The AI Infrastructure Gamble Straining Cash Flows
Oracle’s spending trajectory mirrors the broader hyperscaler race but carries distinctive risks. Unlike Microsoft or Amazon, whose core software margins historically exceeded 30 percent, Oracle is layering industrial-scale power, cooling, and GPU procurement onto a business whose legacy database franchise still accounts for the majority of profits. The company projects up to $70 billion in capital expenditures for fiscal 2027, a figure that would widen the free-cash-flow deficit further before any material contribution from new AI workloads appears on the income statement. Down 48%, Is Oracle Stock a Buy Right Now?
Analysts note that Oracle’s closed-loop water systems and long-term power contracts are intended to stabilize operating costs, yet the sheer scale of GPU clusters required for frontier-model training introduces volatility in both power pricing and hardware depreciation schedules. Investors are therefore discounting near-term margins even as management insists the buildout is fully contracted.
Rising Debt and Market Skepticism
S&P Global’s downgrade of Oracle’s credit rating to BBB-—one notch above junk—explicitly cited the AI infrastructure program as the driver. The rating action followed confirmation of a $300 billion AI computing contract with OpenAI and a separate $260 billion leasing commitment, both of which require front-loaded capital outlays. 3 Wall Street Analysts Have Oracle Going to $400
Credit markets have responded with record CDS levels, forcing Oracle to pay higher spreads on new issuance. While the company retains substantial cash-generation capacity from its installed base, the combination of elevated leverage and multi-year investment commitments has narrowed its margin of safety should AI demand soften or power costs spike.
Record Backlog Underpins Bullish Projections
Offsetting these concerns is Oracle’s remaining performance obligation of $638 billion, which surged 363 percent year over year. Management has guided that $77 billion of this backlog will convert to revenue in fiscal 2027, with an additional 34 percent expected over the subsequent two years. Three sell-side firms—Guggenheim, Jefferies, and Global Equities Research—now carry $400 price targets, implying more than 200 percent upside from recent levels. 3 Wall Street Analysts Have Oracle Going to $400
The backlog growth stems from enterprise customers committing to multi-year OCI capacity reservations tied to specific AI workloads. Because these contracts are binding and often include minimum revenue commitments, they provide greater visibility than the consumption-based models prevalent at other cloud providers.
Expanding Ecosystem Through Partnerships
Oracle has simultaneously accelerated multicloud connectivity. Its Interconnect for AWS service reached general availability in early August, offering customers a private, high-speed, fee-free link between OCI and Amazon’s infrastructure. The move complements last week’s expanded agreement with Alphabet to host Gemini 3.1 Flash Lite and Gemini 3.5 Flash models on OCI, giving enterprise customers direct access to Google’s latest models without leaving Oracle’s environment. Oracle Interconnect for AWS Is Now Generally Available
These partnerships reduce customer friction and increase the stickiness of OCI workloads, yet they also embed Oracle deeper into an ecosystem whose economics remain dominated by GPU suppliers and power providers. Execution risk therefore shifts from demand generation to operational delivery across an increasingly distributed footprint.
Local Realities of Hyperscale Expansion
The physical footprint of this buildout is already generating friction at the community level. In Saline Township, Michigan, Oracle and OpenAI have submitted a permit request to discharge treated wastewater from a planned data center into the Saline River. Residents have raised concerns about long-term environmental monitoring, even though the facility will use a closed-loop cooling system. Saline River at center of permit request
Such permitting battles illustrate how the capital intensity of AI infrastructure now extends beyond balance sheets into regulatory and social license domains. Delays in any single jurisdiction could affect the timing of revenue recognition from the $638 billion backlog.
Oracle’s transformation into an AI infrastructure provider is no longer a theoretical debate; it is a multi-year capital cycle whose outcome will be determined by the pace at which contracted workloads materialize and by the company’s ability to maintain investment-grade access to debt markets. The next earnings release in September will offer the first detailed update on utilization rates and margin trajectories under this new operating model.
