It’s been news in recent hours that Sam Altman is looking for 7 trillion dollars for a project related to chips for Artificial Intelligence. Yet, according to the CEO of NVIDIA – who knows a lot about AI chips – the Sam Altman’s demands are excessive: The Artificial Intelligence sector does not need such large numbers.
According to various press sources, Sam Altman would like to build an “Empire of AI chips”, moving OpenAI from a development purely linked to software to one that also includes hardware for Artificial Intelligence. In particular, OpenAI is expected to compete with NVIDIA, Intel and AMD (the latter also paired with Microsoft, the largest financier of OpenAI itself) in a sector that is becoming increasingly competitive.
So, Altman would like to raise financing for 5-7 trillion dollars (i.e. for a figure between 5,000 and 7,000 billion dollars) to build a series of new Fab think for AI chips. The request of the CEO of OpenAI, however, is extreme, in some ways even impossible to fulfill: at the moment, in fact, the entire AI sector worth a trillion dollars.
Precisely for this reason, the NVIDIA CEO Jensen Huang is skeptical towards the request of the Number One of Open AI. According to Huang, in fact, such a large investment would not be necessary at all to build an alternative production chain to the GPUs currently on the market. Obviously, the statements of the CEO of the Green Team must be taken with a pinch of salt, since NVIDIA dominates the AI sector and has every interest because OpenAI does not enter.
At the World Government Summit, therefore, Jensen Huang declared that “If you think computers aren’t going to get faster over time, you might come to the conclusion that we need 14 planets and three galaxies and four suns to make it all work.”explaining that it will not be so much the quantity of chips produced, but the rate of innovation in the development of GPUs for AI, to drag the computing power of Artificial Intelligence towards previously unseen heights. In such a context, investments like those requested by Sam Altman would be simply enormous, because they would put innovation aside to focus simply on mass production of chips.
Indeed, Huang concluded his reasoning by saying that “we must remember that the performance of the architecture will improve over time, therefore you can’t just think about buying more computers. You also have to assume that computers will get faster, and therefore the overall number of calculators you will need will not be as high.”.