Training of the next generation AI models worsens the problem. In addition to performing existing AI models such as those that chatgpt, OpenAI is constantly working on new technology in the background. It is a process for which thousands of specialized chips are continuously running for months.
The circular investment question
The financial structure of these deals between OpenAi, Oracle and Nvidia has started the research of observers in the industry. Earlier this week, Nvidia announced that it would invest up to $ 100 billion, because OpenAi Nvidia systems uses. As Bryn Talkington from the required capital management told CNBC: “Nvidia invests $ 100 billion in OpenAi, who then returns OpenAi and returns it to Nvidia.”
Oracle's scheme follows a similar pattern, with a reported $ 30 billion deals per year in which Oracle builds facilities that Pays OpenAi for use. This circular current, where infrastructure providers invest in AI companies that become their largest customers, has raised eyebrows about whether these real economic investments or extensive accounting maneuvers.
The schemes are becoming even more complicated. The information reported this week that Nvidia discussed about leasing his chips to OpenAi instead of selling them downright. Under this structure, Nvidia would create a separate entity to buy its own GPUs, and then lease them from OpenAI, which adds another layer of circular financial engineering to this complicated relationship.
“Nvidia seed companies and gives them the guaranteed contracts needed to increase debts to buy GPUs from Nvidia, although these companies are terribly unprofitable and will eventually die from a lack of real question,” wrote Tech critic Ed Zitron about Bluesky last week. Zitron referred to companies such as Coreweave and Lambda Labs, who have collected billions in debts to buy Nvidia GPUs, partly based on contracts from Nvidia itself. It is a pattern that reflects the OpenAI arrangements with Oracle and Nvidia.
So what happens when the bubble jumps? Even Altman himself warned last month that “someone will lose a phenomenal amount of money” in what he called an AI bubble. If the AI requirement does not meet these astronomical projections, the massive data centers built on physical soil will not just disappear. When the DOT-Com Bubble burst in 2001, the fiber optic cable that was laid during the tree years found use as an internet demand. Similarly, these facilities can possibly run to cloud services, scientific computer use or other workload, but what might be huge losses for investors who have paid AI-boom prices.
