Forget Nvidia for a second. If you want to understand how far this AI cycle has pushed corporate balance sheets, look at Oracle.
Oracle’s free cash flow to price ratio went negative – deeply negative – sitting at -1,179 times in fiscal 2025 and still deeply negative at -27.5 times in the most recent fiscal year. Read that twice. A company with a market capitalization of $407.2 billion, trading at $141.38 a share, is generating essentially no free cash flow relative to its valuation right now – because it is plowing capital into AI data center capacity at a pace that outstrips what its core software and cloud business throws off in cash.
This is not subtle. This is a company that decided, apparently at the board level, that the AI infrastructure land grab was worth funding with debt and negative near-term free cash flow rather than waiting for organic cash generation to catch up. That’s a real bet, with real balance sheet consequences, and it deserves to be discussed with the seriousness it warrants rather than folded into generic “AI winners” narratives.
Here’s what makes Oracle genuinely interesting rather than simply alarming: the trailing P/E sits at a reasonable-looking 18.53 times, and even the forward-looking figure is only 38.7 times – multiples that would look defensible for a legacy enterprise software business with Oracle’s customer stickiness, assuming the capex actually converts into contracted, recurring cloud revenue at the margins the company is promising. Gross margin has compressed from 70.5 percent to 65.8 percent year over year, which is exactly what you’d expect to see if a company is scaling low-margin infrastructure capacity ahead of the higher-margin software revenue that’s supposed to run on top of it eventually.
The entire investment case for Oracle right now is a timing bet: does the contracted cloud backlog convert to cash flow before the debt service and depreciation load becomes unsustainable? If it does, today’s negative free cash flow reads, in hindsight, exactly like Amazon’s AWS build-out looked in the mid-2010s – painful in the moment, obviously correct in retrospect. If it doesn’t – if hyperscaler demand for third-party cloud capacity plateaus, or if Oracle’s specific customer concentration in that backlog turns out to be thinner than advertised – this is a textbook case of a company that levered up at the top of a cycle.
What’s notable is how little public debate there is about which scenario is more likely. Compare the volume of commentary on Nvidia’s valuation to the volume of commentary on Oracle’s balance sheet trajectory, and the imbalance is striking given that Oracle’s bet is arguably the more consequential one for its own shareholders. Nvidia can disappoint and still be a wildly profitable, dominant business. Oracle’s AI bet, if it goes wrong, hits a company that’s using leverage most of its historical enterprise software cohort never touched.
I’m not calling this a mistake. Big bets funded with debt against a genuine platform shift have made fortunes before. I’m saying the risk is concentrated and specific in a way the market hasn’t fully priced into the conversation, even if the multiple itself looks unremarkable on paper.
