The next generation models are exacerbated by the problem. In addition to the launch of existing artificial intelligence models, similar to those that feed Chatgpt, Openai constantly working on new technologies in the background. This is a process that requires thousands of specialized chips to work continuously for several months.
Round investment issue
The financial structure of these transactions between Openai, Oracle and NVIDIA conducted a thorough study from industry observers. Earlier this week NVIDIA announced It invests up to $ 100 billion because Openai deploys NVIDIA Systems. Like BRIN Talkington of Requisite Capital Management said CNBC: “NVIDIA invests in Openai in the amount of $ 100 billion, which Openai then turns back and returns it Nvidia.”
Oracle arrangement should A similar scheme with an agreement on 30 billion dollars per year, where Oracle builds funds that Openai pays for use. This circular flow, which includes infrastructure suppliers investing in the company of artificial artificial art, which become their largest customers, have Raised eyebrows About whether they represent genuine economic investments or complex accounting maneuvers.
The agreements become even more confusing. Information reported This week, NVIDIA discusses the leasing of its chips on Openai, and not sell them. In accordance with this structure, NVIDIA will create a separate organization to buy its own graphic processors, and then rents them to Openai, which adds another level of circular financial engineering to these difficult relations.
“NVIDIA SEEDS Companies gives them guaranteed contracts necessary to increase the debt to buy graphic processors from NVIDIA, even if these companies are terribly unprofitable and will ultimately die from the absence of any real demand.” wrote Technological critic Ed Qitron on Bluesky last week about an unusual stream of investment in the infrastructure of AI. Zitron referred to companies such as Coreweave and Lambda Labs, which increased billions of debt to purchase NVIDIA graphic processors, partially based on contracts from NVIDIA itself. This is a sample that reflects OpenAi arrangements with Oracle and Nvidia.
So what happens if a bubble appears? Even Altman himself warn Last month, “someone will lose the phenomenal amount of money” in what he called the AI bubble. If the requirement of AI does not meet these astronomical forecasts, mass data centers built on physical grounds will not just disappear. When in 2001 the bubble burst out of the ground, the fiber-optic cable laid during the years of the boom, ultimately found that the demand on the Internet insisted at home. Similarly, these objects can potentially plunge into cloud services, scientific calculations or other workloads, but despite the fact that it can be mass loss for investors who paid prices for AI-BOOM.
