Oracle's AI Bet: Boom, Doubt, and the Signals Finance Teams Can't Ignore
Oracle's 2025 run encapsulates the core debate on AI: massive upside vs. mounting risk. The company rode early-year enthusiasm on a $500 billion US AI infrastructure joint venture with OpenAI and SoftBank - "Stargate" - announced at the White House, sending shares higher and briefly making Larry Ellison the world's richest person.
Then the narrative flipped. Debt-fueled spending, widening credit risk signals, and concentration tied to a single anchor tenant have turned Oracle into a case study for AI bubble fears.
The setup: AI-fueled surge, sky-high expectations
Following June and September earnings, AI-driven deals were positioned to push Oracle's cloud segment revenue to $166 billion by 2030. Momentum was clear, enthusiasm was strong, and capital markets rewarded the story.
But stories need cash flow - especially when the bill is this large.
Funding the buildout: leverage up, CDS flashing
Investors started questioning the rising reliance on debt across Big Tech to bankroll AI infrastructure. Credit default swaps - insurance-like contracts on corporate default risk - began trading even for the highest-rated tech names, signaling a shift in how credit markets are pricing risk. Here's a quick refresher on CDS.
"As these firms such as Oracle have issued more debt, they've become more leveraged, which from a credit perspective means they're riskier," S&P Global Market Intelligence analyst Gavan Nolan said. Oracle issued nearly $26 billion in bonds in 2025; its five-year CDS cost hit the highest level since 2009.
Balance sheet tension: debt, cash flow, and hidden commitments
In the latest results, total debt rose 40% year over year to $124 billion, while cash outflow climbed from $2.7 billion to $10 billion. After earnings, Oracle disclosed via an SEC filing $248 billion in additional lease commitments beginning in fiscal Q3 2026 through 2028 - mostly data centers - not reflected on the balance sheet. You can monitor company disclosures via EDGAR.
Morningstar's Luke Yang noted Oracle now has "very little room for error." He added that free cash flow is more constrained than hyperscalers like Microsoft and Alphabet, which limits funding flexibility if execution slips.
Concentration risk: OpenAI as anchor tenant
Oracle's remaining performance obligations are heavily tied to OpenAI - at least $300 billion - a detail that pressured shares once revealed. The market is questioning whether OpenAI can hit aggressive revenue targets amid competition from Google.
OpenAI's costs are projected at $1.4 trillion over the next eight years across infrastructure deals with Nvidia, CoreWeave, AMD, Broadcom, and Oracle. That scale has investors on edge. Hedgeye's Andrew Freedman summed it up: the demand may not materialize at the size and speed needed to justify the spend.
Execution watch: build timelines, counterparty health, leadership change
Oracle says it can repurpose AI infrastructure quickly if a client can't pay. Still, timing matters. Bloomberg reported Oracle pushed some OpenAI-tied data centers to 2028 from 2027; the company denied delays. The Financial Times reported Blue Owl wouldn't back a large Michigan facility; Oracle said a different equity partner will finance it.
There's also leadership turnover: co-CEOs Clay Magouyrk and Mike Sicilia stepped in as Safra Catz moved to the board. Governance continuity and decision speed will be watched closely as capital intensity rises.
Investor sentiment split
"Oracle has become the poster child for fears of an AI bubble," investor and tech analyst Cory Johnson said. Even so, bulls lean on execution history: "Oracle has historically delivered on their promises… Every year and every decade, Oracle ends up ahead, and their competitors fall behind."
Oracle declined to comment.
What finance teams should monitor next
- Credit risk signals: CDS spreads, bond yields, and market depth for new issuance.
- Leverage and liquidity: net debt trajectory, cash burn vs. operating cash flow, refinancing needs.
- Capex and leases: timing, off-balance-sheet commitments, and any changes to data center delivery schedules.
- Backlog quality: concentration in OpenAI, conversion rates, and contract terms that protect Oracle if clients stumble.
- Utilization and monetization: time-to-revenue from new capacity, pricing, and margins on AI workloads.
- Counterparty health: OpenAI's funding, competitive pressure, and its ability to translate usage into revenue.
- Management updates: CEO cadence, guidance discipline, and capital allocation signals.
Scenarios to frame your models
- Bull case: Data centers come online on plan, utilization ramps, OpenAI performs, and Oracle diversifies demand. Leverage stabilizes as free cash flow improves.
- Base case: Mixed execution with modest delays, steady but uneven AI demand, and gradual improvement in backlog conversion. Credit spreads stay elevated but contained.
- Bear case: Slower AI monetization, client delays or renegotiations, and higher build costs extend payback periods. Spreads widen, refinancing gets pricier, and capex/lease load pressures cash flow.
Practical takeaways
- Re-rate your risk for concentration: RPOs dominated by one counterparty change the profile of the backlog.
- Separate story from cash: watch free cash flow coverage of capex and leases, not just revenue targets.
- Let credit markets guide you: if CDS keeps widening, model higher financing costs and tighter covenants.
- Build sensitivity around utilization timing: a year's delay on ramp can swing outcomes more than headline growth assumptions.
If you're benchmarking practical AI use cases and ROI inside finance teams, here's a curated list of tools to pressure-test against your models: AI tools for finance.
Your membership also unlocks:
