Surrounding Artificial Intelligence Investment in and trading of computing infrastructure has become a core driver of the current US stock market. However, this frenzied trend is facing increasing skepticism, even from Microsoft. CEO Satya Nadella also indicated that there might be problems.
In an interview, Nadella and OpenAI CEO Altman discussed the current state of the artificial intelligence industry. Nadella stated that computing power is not the current bottleneck for industry development; there is a lack of sufficient electricity and data centers. Space is the obstacle.
He emphasized that there are currently a large number of AI chips in the industry that are stuck in inventory and cannot be used . Microsoft 's problem is not the supply of chips, but whether it can complete the infrastructure construction quickly enough and bring its data centers close to optimal performance.
When asked whether this backlog would lead Microsoft to reduce its orders from Nvidia, When asked about purchasing GPUs, he responded that short-term demand was difficult to predict and the final outcome would depend on the supply chain situation.
The energy consumption of artificial intelligence has been a key point of contention surrounding the industry's development, especially given the increasingly strained traditional power supply. The center is accelerating the development of its own small modular nuclear reactor to expand its energy supply capacity.
OpenAI has even called on the U.S. federal government to build 100 gigawatts of power generation facilities annually, calling it a strategic asset that will determine the international race in artificial intelligence.
Beyond seeking growth in electricity supply, some companies are exploring feasible ways to reduce the energy consumption of large models. Altman pointed out in the interview that OpenAI will one day create a consumer-grade device that can run GPT-5 or GPT-6 models locally with extremely low power consumption.
However, this idea contradicts the massive investment required for data centers , because the extremely low power consumption and local operation of the model mean that the widespread construction of large data centers is not necessary, which could pose a significant risk to the artificial intelligence industry.
Data shows that since the beginning of this year, the issuance of bonds related to artificial intelligence has exceeded $200 billion, with most of the funds going towards building computing infrastructure. However, analysts warn that high industry concentration coupled with concerns about capital sustainability could trigger systemic risks.
(Article source: CLS)
