Oracle is beginning to turn into one of those cautionary case studies. And that’s not good news for healthcare. From their faltering EHR to the hacking of their software, Oracle’s news hasn’t been cheerful. While still profitable in the quarterly reports, the debt load and its obligations in their chosen wrenching changeover from a software developer/seller to an AI infrastructure landlord have been followed by speculation. The latest speculation from June is selling their EHR business [TTA 2 June]–if they can find a buyer. The EHR is a distant second to Epic in every metric [TTA 27 May] and Oracle has an immense obligation to the Federal Government for both the Military Health System (MHS) and VA’s EHR Modernization that won’t complete projected rollout until 2031.

A recent analysis published in Seeking Alpha (may require free account, PDF here) will give one further pause.

The debt load is a crusher. There’s anywhere from $72 to $100 billion in debt on Oracle’s books, and that preceded the PIMCO financing of $16.3 billion that included equity. Their credit rating from major raters is one step above junk: BBB-negative from S&P Global Ratings and Moody’s Baa2 Negative outlook. [TTA 7 May] Oracle has already projected it needs more: an additional $40 billion in fiscal 2027 through a mix of debt and equity.

Net debt is soaring. Oracle ended FY 2026 with net debt of $97.6 billion. This represents a ~20% increase versus the $81.4 billion of FY 2025. Looking forward, this analyst is projecting that the net debt by end of FY 2027 will be in the range of ~$120 billion.

There’s another little problem called cash flow. Years of positive free cash flow, even during the pandemic recession, turned sour in their FY 26 to a negative $23.7 billion. The analysis states that Oracle is now heavily dependent on external cash–debt–to finance its AI infrastructure building. S&P also noted that capital expenditure has been geometrically increasing: $6.9 billion (2024), $21.1 billion (2025) and $55.7 billion (2026). The guidance for fiscal 2027? Between $90 and $95 billion. That is over a 70% increase. Cash flow clearly can’t pay down the debt they already have, a fact that banks factored in to cut off the lending spigot.

Its performance obligations are also crushing. $553 billion has been reported for OpenAI alone [TTA 7 May].  This article reports remaining performance obligations of $638 billion, about half of which is for OpenAI. Despite the disparity in numbers (perhaps completed obligations?), in this analyst’s view, something happens to OpenAI–for instance, a pullback in the confidence of markets to finance their aggressive expansion (and, this Editor would add, an IPO)–Oracle would be hit hard. To paraphrase an overworked analogy, if OpenAI stumbles and sprains its ankle, Oracle would break both legs.

But..but…isn’t demand supposed to be increasing, infinitely?  Increased demand is critical to continued financing. Yet there are leading indicators that demand is leveling off. Corporations are increasingly limiting internal use of AI by employees to tamp down soaring token usage costs and control budgets. These companies have also found that AI tools often do not work as well in replacing humans as they were cracked up to be. And now seemingly every software offering is ‘powered by AI’. The newness is wearing off.

The chip cycle is speeding up–not good news for AI infrastructure builders.  Nvidia is now in full production of its Vera Rubin chip, which is dramatically faster than the previous generation Blackwell chip. Our March reporting picked up that Oracle had forward commitments with Nvidia to use the older Blackwell chip for the OpenAI data centers, and OpenAI wasn’t happy about that. It’s not a simple matter of going to Nvidia and returning old chips and getting new ones–that isn’t done. Vera Rubin is also twice as expensive on a per-rack basis than Blackwell–$7.8 million versus $4 million. Add to this the long taffy pull of building (online is 12-18 months minimum), power, water, and labor costs…it adds up, as any builder will tell you.

Given the above factors, it shouldn’t be any surprise that Oracle stock is down 30% since last December and continues to decline.

Extra: The article also contains a short history of financial cycles. There are parallels drawn comparing the AI boom we are in to similar technology-driven boom and bust cycles, dating back to the building of canals in the 1700s, railways in the 1800s, and electrification in the early 1900s.  

Editor’s note: Any analysis of Oracle’s financial health is based on this article, previous articles, multiple sources, and Oracle’s disclosures. Our interest is primarily in the well-being of Oracle Health and the markets it serves. We do not engage in stock picking and none of the above should be construed as financial advice.