The $99 Billion Mirage
Oracle just dropped its Q2 2026 earnings report on December 11, and the market is finally waking up to the smell of burning cash. While Larry Ellison spent most of the call talking about sovereign AI and giant data centers, I spent my morning digging through the latest 10-Q filing. The numbers tell a story that the press release tried to bury. Oracle claims a Remaining Performance Obligation (RPO) of $99 billion, a staggering 18 percent increase. But here is the catch. This is a backlog, not a bank account. My analysis shows that the actual conversion rate of this backlog into realized revenue is slowing down. We are seeing a massive gap between what Oracle promises to build and what enterprises are actually paying for today.
Capex is Cannibalizing the Core
The cost of staying in the AI race is becoming unsustainable. Oracle has committed to spending nearly $15 billion on capital expenditures this year. This is not just money for new servers. It is a desperate attempt to keep pace with Microsoft and Amazon. I find it suspicious that while Oracle Cloud Infrastructure (OCI) revenue grew at 25 percent this quarter, their operating margins are being squeezed by the sheer cost of Nvidia H200 clusters. The “Nvidia tax” is real, and Oracle is paying it in blood. The stock, currently hovering near $188, is priced for perfection, but the foundation is starting to crack under the weight of these infrastructure costs.
Salesforce and the Death of the Seat
Oracle is not the only one facing a reckoning. I have been tracking Salesforce (CRM) and their recent push into “Agentforce.” The pivot is telling. For twenty years, Salesforce grew by selling licenses per human user. Now, they are admitting that AI agents will do the work of humans. This is a direct threat to their legacy business model. If a company needs 50 percent fewer customer service reps because of AI, they need 50 percent fewer Salesforce seats. Per Reuters reports from earlier this week, enterprise software spending is shifting from “applications” to “compute.” This is a secular decline disguised as a technological upgrade. Salesforce is trading at $312, but their forward guidance suggests a growth wall that no amount of AI marketing can climb.
The Adobe Problem: Generative Erosion
Adobe (ADBE) is often cited as an AI winner, yet their price action tells a different story. Since the release of advanced video generation models in late 2024 and throughout 2025, the moat around Creative Cloud has evaporated. I spoke with three creative agencies last week that are shifting budgets from Adobe subscriptions to direct API access for open-source generative models. Adobe is forced to integrate these features just to stay relevant, but they cannot charge the premium they once did. The market is ignoring the fact that AI has commoditized the very tools Adobe spent decades perfecting.
A Fragmented Market Landscape
The following table illustrates the disconnect between current valuation and the actual growth in free cash flow (FCF) for the major software players as of mid-December 2025.
| Ticker | Price (Dec 17) | P/E Ratio | FCF Growth (YoY) | The Reality Check |
|---|---|---|---|---|
| ORCL | $188.45 | 34.2x | -2.1% | Capex is eating the cash flow. |
| CRM | $312.10 | 29.5x | 4.4% | Seat-based model is dying. |
| ADBE | $538.90 | 27.1x | 1.2% | Generative AI is a margin killer. |
| MSFT | $425.15 | 31.0x | 8.8% | Azure is the only real winner. |
The Hidden Risk of Sovereign AI
Larry Ellison loves the term “Sovereign AI.” He claims every nation will need its own Oracle-powered cloud. I find this narrative flawed. Most nations do not have the power grid capacity or the localized data to make this a reality by 2026. This is a long-term play being used to justify short-term stock price inflation. According to the latest Bloomberg data on global data center energy constraints, the bottleneck for AI growth in 2026 will not be software, it will be electricity. Oracle is selling a dream that the utility companies cannot support.
Watch the March 2026 Milestone
Investors should stop looking at the AI hype and start looking at the debt maturity schedules. Oracle has significant debt coming due in the next eighteen months, and with the 10-year Treasury yield stubbornly high at 4.2 percent, refinancing will be expensive. The next critical data point is the March 2026 earnings release. If the RPO conversion does not accelerate by then, the $188 price floor will vanish. Watch the net debt-to-EBITDA ratio. If it crosses 3.5x in the next quarter, the AI narrative will not be enough to save the credit rating.
