$Oracle(ORCL.US) Some people say Oracle has financial problems because free cash flow is very important. Oracle's revenue isn't as high as a few cloud giants, and it carries heavy debt, so it's likely to be the first company to blow up. My view is that cash flow is certainly important, but a detailed distinction is needed. Compared to the other companies, Oracle indeed carries higher risks, but it's far from the point of potential debt default or bankruptcy. First, if cash and equivalents are sufficient to cover interest payments, and future revenue and profit growth expectations are enough to cover long-term debt, then the risk is manageable. It's like if a person's income is high and stable enough, they can take out a loan to buy stocks that are expected to rise in the long term, as long as the expected rise is significantly greater than the interest. Oracle currently has three main concerns: 1. High customer concentration, 2. Data centers may become worthless in the future due to changes in development paths, 3. Excess computing power. The latter two are just ghost stories, easily debunked. Currently, OPENAI does seem to be overshadowed by Anthropic, and it's hard to see how it could afford so much financially. But there are four points to note: First, the capabilities of open-source models are getting stronger, which will impact Anthropic, which mainly focuses on 2B, while the potential impact on OPENAI, with its 1 billion active users, is significantly smaller. Second, OPENAI previously worked on many side tasks, leading to poor performance in programming models, but it is now clearly catching up and even surpassing Claude in specific areas. Third, the US government clearly has a better relationship with OPENAI, and historically, the US government has always encouraged competition, so there is motivation to restructure and provide more policy support to OPENAI. The previous Fable model was under review for a longer time than GPT5.6, and even after release, many of its responses were restricted, which is evidence. Fourth, Ellison is very close to Trump personally, and Trump himself holds a significant position in Oracle. Trump also has ample motivation to support Oracle in policy details.
Finally, and perhaps most importantly, please remember that AI has just begun. It is currently in a stage where its capabilities are already very strong, but people still don't know exactly how to use it. AI is far from a zero-sum, red-ocean game. It's about making the pie bigger so everyone can make money. Currently, computing power costs are still relatively high, leading many companies to impose token fee limits on employees. Besides semiconductors being expensive themselves, the high cost of computing power is more importantly due to the model's architecture and people not knowing how to reasonably allocate work and utilize tokens. These problems will be solved over time. Once computing power costs come down, more people will definitely use AI. In other words, AI is currently precisely because costs are too high, and people haven't figured out more efficient ways to use it, so demand is still suppressed. However, even with this suppressed demand, it's still visibly a state of computing power shortage. In the future, once computing power costs drop and people become more adept at using AI, the unleashed demand for AI will be almost endless. But the data centers and electricity that can be built on the ground are limited. Under such circumstances, do people still think there will be a surplus of computing power in the future?
