The CEO of Nvidia believes that, unlike the dot-com boom two and a half decades ago, there is real demand behind the AI craze.

The similarity between AI and the dot-com bubble stems not from the underlying technologies and their impact on industry but from stock valuations and the exorbitant retail profits people experienced in the past. Nvidia’s years of growth, which broke through stock valuation barriers in a matter of months, led to this conclusion. However, Huang has called AI a bubble and believes that, unlike in the dot-com era, the introduction of “dark fiber” has created real demand for AI computing power.

When asked if he was concerned that people would reach a point where they did not understand the technology, acted hastily, and caused the bubble to burst, he responded, “If not, what is the biggest misconception about demand for AI that differs from the dot-com era? If not, what is the biggest misconception about the demand for AI that differs from the dot-com era? Huang said that, during the dot-com era, the vast majority of installed threads were dark. This means that the industry installed far more threads than it needed. Today, almost every GPU is turned on and in use.

“Dark fiber” refers to the massive expansion of internet infrastructure from the mid-1990s to the end of the decade. Companies expected a dramatic increase in demand for fiber optic cables. Consequently, service providers dug trenches and laid more cable than necessary to future-proof their operations. However, their assumption was incorrect, and the return on investment was virtually zero. Thus, the term refers to the creation of artificial demand.

Huang claims that AI has undergone tremendous development, yet mainstream consumers still consider ChatGPT and image-generating applications to be the pinnacle of this technology. AI has evolved to the point where it can effectively think and base itself on research. However, this technology has not yet spread to the masses. According to Huang, companies’ computing needs are growing exponentially and the number of queries is increasing, indicating that demand is real.

AI has enormous growth potential and requires computing capacity, regardless of whether it is provided by Nvidia, AMD, or Intel. However, with such a massive rollout, limitations must be considered, such as power supply issues. Another important consideration is whether large companies will be able to integrate AI chips manufactured by Nvidia or other companies.

Source: WCCFTech