Oracle investment based on ChatGPT advice costs €25,000

€25,000 invested in Oracle because ChatGPT said so

An investor opens a €25,000 position in Oracle after using a detailed prompt to ask ChatGPT for undervalued stocks. He uses CFDs rather than physical shares and admits: "I'm opening this thread to cover myself in glory or let you laugh at me for losing money by trusting AI." The stock was trading near $146.49, far from its September 2025 peak of $350. While consensus remains Buy, the market is split between Oracle's massive AI backlog and the cost of financing it.

The analysis behind the decision

The investor reproduces the report generated by the model: tables of forecasts and target prices, a bearish scenario, and a probability map. The summary highlights two facts: the value dropped in December due to weak guidance and high capex/debt, then rebounded in March after RPO (remaining performance obligations) jumped to $553 billion, with projected revenues of $90 billion for 2027.

This is the core thesis. Oracle is no longer just a database company: it is an infrastructure bet on AI with growing debt. Buying it means buying that risk, not a stable dividend.

The 66% drop and the dot-com bubble ghost

The most repeated data point is the crash. Oracle fell from $350 to $145, a 66% drop from highs. During the dot-com bubble, it fell 82%, and some argue this is the second worst decline in its history. It proved its resilience back then, defenders say: while many peers vanished, Oracle survived, recovered, and grew.

Opponents argue simply: a big drop doesn't make a stock a bargain. Some warn it could stagnate for years. Others note that in a major market correction, Oracle could hit $80, and since CFDs are used, losses are amplified if the price turns against him.

The leverage problem: €25,000 isn't really €25,000

Here lies the key technical issue. With CFDs, actual capital committed is a fraction of the nominal amount. One participant estimates effective exposure at around €5,000 and warns that such leverage only makes sense with strong upward momentum and tight stop-losses, as a reversal triggers margin calls quickly. The investor later admits his entry was poor: he bought at $146 during a bounce, when he could have waited for $120.

Closing the trade and re-entering

He closes the position. "I think it was a stupid and excessive entry. I don't feel comfortable with €25,000 in one stock." Shortly after, with the stock at $138, he says he feels tempted to throw coins at it but would prefer to re-enter near $120. He ends up putting in €1,500 at $136.46. From the initial nominal amount to a residual position of €1,500 in days.

AI bias and the underlying debate

The main criticism isn't about Oracle, but the method. One user notes that asking ChatGPT for five diversified stocks resulted in five AI plays; asking for non-AI stocks still yielded three AI-related ones. The conclusion is that the model has a thematic agenda and fails at complex decisions. Another points out that models are often configured to explain why their own investment advice shouldn't be trinc blindly.

The background is familiar: debt-to-equity over 400%, P/E above 25, and a long-term downtrend. Bulls argue that if AI succeeds, Oracle will reach a trillion-dollar market cap as a new megacap. Bears reply that everyone said that in 1999.

---

With the position closed and €1,500 remaining at $136.46, the unanswered question is whether the problem was Oracle or the prompt. No one in the conversation managed to separate the two.