To put it mildly, the concept is grandiose and would completely upend the current semiconductor business.
The CEO of OpenAI is reportedly working on a plan to change the semiconductor industry, as stated in a new report that was published by the Wall Street Journal (payment required). By securing up to seven trillion dollars in investment, the CEO of the business that is responsible for the industry-changing ChatGPT intends to establish a new worldwide network of foundries and power plants that will generate individualized silicon that will be utilized for artificial intelligence. It is a daring and ambitious concept that, if it ever comes to reality, will completely transform the landscape of the semiconductor business. However, this is a very large if.
According to the article, Altman has already started talking to foundries; however, it is not obvious which foundries would be interested in accepting such an offer. It has been reported by Tom’s Hardware that the research makes reference to TSMC; however, it does not name Intel, Samsung, or Global Foundries. According to reports, Altman seeks to establish a global network of fabrication facilities in order to manufacture artificial intelligence chips for OpenAI and other businesses. In order to accomplish this, he is going to require between five and seven trillion dollars, which is a sum that appears to be tough to get, even from a large number of different sources.
The amount of money also exceeds the current spending of the business; according to Tom’s Hardware, the semiconductor sector is anticipated to become a trillion-dollar industry, although this goal is not likely to be reached until the year 2030. Considering that it is almost half of that amount in its existing state, it is not obvious why Altman believes that such a large amount of money will be required. It’s possible that he has a mental picture of all the costs that are connected with establishing numerous fabs, including the infrastructure that supports them. On the other hand, it is predicted that even fabs built from the ground up, like as Intel’s new fab in Ohio, will only cost $100 billion over the course of their lifetime. The next memory fab that Micron is planning to build in next York will have a price tag of one hundred billion dollars.
What remains to be seen is the extent to which Altman will be successful in this endeavor of great ambition. According to reports, it is currently only existing in the form of conversations between Altman and potential investors and chip manufacturers. Nevertheless, it is an original concept that would require a significant amount of time to materialize. It also appears to be a hazardous investment in light of the ongoing competition among TSMC, Intel, and Samsung to diversify the locations of their fabrication facilities in order to safeguard themselves against geopolitical instability and potential disruptions to supply chains in the future. Furthermore, on a much smaller scale, a global network of fabs that produce specialized artificial intelligence silicon sounds like it could be similar to what Nvidia intends to do. The fact that it is already utilizing Nvidia technology allows us to speculate on whether or not it will also be a part of Altman’s plans at some time in the future.