Oracle’s $500 Billion AI Bet vs. $100 Billion Debt Pile: Earnings Could Make or Break This “Backlog Bubble”

Oracle’s earnings on Wednesday aren’t just another AI print; they’re basically a referendum on whether a $500 billion capex dream, a $455 billion backlog, and a $100 billion debt stack can peacefully coexist on one balance sheet without giving fixed-income desks a nervous breakdown.

Analysts expect the headline numbers to look solid. Street models generally line up with management’s guidance for mid-teens revenue growth and another strong quarter of cloud momentum, with total revenue for fiscal Q2 expected to grow roughly 14–16% year-on-year and cloud revenue up in the low-to-mid 30% range. Non-GAAP EPS is expected to land around management’s guide of roughly $1.61–$1.65, with at least one large broker (Jefferies) modeling EPS slightly ahead of consensus. The problem, as CFRA’s Angelo Zino points out, is that “I don’t really think the numbers really matter here this quarter” – the stock will trade on the story around debt, backlog quality, and OpenAI exposure, not a penny or two of EPS.

That story starts with the backlog. Oracle’s remaining performance obligations (RPO) exploded to $455 billion in Q1 FY26, up 359% year-on-year and more than tripling in a single quarter, with management signaling that RPO should “exceed $0.5 trillion” as additional mega-deals are signed. A huge chunk of that – around $300 billion – is tied to OpenAI’s multi-year compute commitments. As Zino notes, that implies a “hockey stick type of inflection” that really kicks in beyond FY27, with OpenAI’s ramp meant to drive a 9x increase in OCI run rate by decade-end. In the near term, there is “no concern about the demand trajectory,” but the market is increasingly skeptical that those long-dated commitments will be fully monetized, especially if Gemini’s recent momentum eats into OpenAI’s share.

That skepticism is amplified by Oracle’s balance sheet. As of Q1 FY26, Oracle had about $10.4 billion in cash and equivalents and roughly $24.6 billion in total current assets. On the liability side, it carried around $91 billion of notes payable and other borrowings (current plus non-current), total liabilities north of $155 billion, and equity of roughly $24.7 billion – implying net debt of roughly $100 billion and a net debt-to-equity ratio just over 4x.(Oracle Investor Relations) Operating cash flow over the trailing four quarters was about $21.5 billion, but capex has exploded: trailing 12-month capex is running roughly $27.4 billion, leaving trailing free cash flow at about negative $5.9 billion.(Oracle Investor Relations) In plain English: Oracle is deliberately running negative free cash flow to build GPU-rich data centers at unprecedented scale, betting that contracted AI revenue will eventually bail out the balance sheet.

Recent financing moves underline that bet. Oracle sold roughly $18 billion of bonds earlier this year, including $3.5 billion of 30-year paper that has since dropped about 8% from its peak as tech credit spreads widened. Project-level financing is also swelling: banks have lined up roughly $38 billion of secured loans to build Oracle-tied AI data centers in Texas and Wisconsin via Vantage Data Centers – the plumbing behind the much-discussed “Stargate” build-out for OpenAI capacity. BNP Paribas estimates Oracle will need “only” an additional $25–30 billion of debt on top of that $18 billion issuance to fund the current build plan, far below the $100 billion-plus worst-case fears swirling in the equity market, but still enough to keep credit desks nervous.

You can see that nervousness in CDS. Oracle’s 5-year credit default swaps have roughly doubled in recent months from about 40–55 bps to the 100–110 bps area, implying a risk-neutral five-year default probability in the high-single digits – elevated for Big Tech, but still consistent with investment-grade ratings. Zino’s take is that CDS has become “a macro instrument as a hedge against an AI crash,” rather than a direct signal that Oracle is about to blow through its covenants. But he also expects free cash flow to remain negative for “the next two to three fiscal years,” with annual net outflows of $10–15 billion as capex heads toward $80–90 billion by the late 2020s.

That’s why this call is really about narrative management. Zino frames the key ask from investors very simply: can management credibly say, “We’ve raised X amount of debt, we generate Y from operations, and for the next 18–24 months we don’t need to tap the capital markets again”? In his view, that alone could be enough to stabilize sentiment and trigger a relief rally in a stock that’s down roughly a third from its September peak and trading at Microsoft-like multiples despite negative free cash flow and much higher leverage. The flip side is ugly: if Oracle can’t define a clear funding roadmap – including how much will come from vendor financing and capital leases versus straight debt – CDS could widen further and the equity multiple compress again.

Investors will also be laser-focused on concentration risk. If the “worst-case scenario” for OpenAI plays out – Gemini cements a durable lead, OpenAI’s revenue trajectory falls well short of the trillion-dollar spend narrative, and some contracted capacity goes unused – Oracle has to prove it can backfill those data centers with other hyperscale and enterprise demand. The company has already talked up large deals with Meta, xAI, and other AI labs, but Zino warns that the Street’s long-term EPS targets “bake in” a very robust OpenAI outcome. Any suggestion that Oracle’s $300 billion OpenAI contribution to RPO is less firm or slower to monetize than implied would hit both the backlog story and the credit story simultaneously.

Against that backdrop, here’s what matters most on Wednesday:

• Bookings and RPO: Not just the headline RPO figure, but the breakdown of how much is tied to OpenAI vs. other AI labs and enterprise customers, and how much is truly “contractually firm” versus contingent on deployment milestones.
• Capex and financing: Updated capex guidance beyond the current ~$35 billion fiscal-year plan, details on the mix of vendor financing, leases, and debt, and any commitment not to re-enter the bond market for a defined period.
• Free cash flow path: A clearer bridge from today’s negative free cash flow to positive territory, including timing, margin assumptions on AI infrastructure, and how quickly new capacity becomes gross-margin and cash-flow accretive.
• Credit and ratings: Management’s stance on preserving investment-grade status, and whether they explicitly address CDS widening and recent downgrades/dovish comments from credit analysts.
• AI demand and competitive positioning: Evidence that AI infrastructure demand remains “insatiable” beyond OpenAI – including Meta, xAI, and other labs – and that Oracle’s differentiated network and database stack can sustain premium economics rather than devolving into a commodity GPU-hosting business.

If Oracle can deliver a quarter that pairs solid numbers with a credible funding and risk-management roadmap, this could mark the point where the stock stops trading as a pure “OpenAI leverage short” and starts trading again as a long-duration AI infrastructure play. If the narrative wobbles, the market may decide that when someone says they’ve “sold 4 million homes to buyers with no income,” it’s time to ask a few more questions before funding the next subdivision.