The CEO of OpenAI plans to build the world's largest AI chip empire, with an estimated investment of 7 trillion yuan.
OpenAI CEO Sam Altman is a man of big ambitions. Not only did he spearhead the recent rise of generative AI, but he also had an even more ambitious goal: to reshape the global semiconductor industry and create a trillion-dollar infrastructure of AI chips.
Altman is pushing for a project to increase global chip production capacity and is in talks with several investors, including the United Arab Emirates, as reported by Wall Street**. Altman may need to raise between $5 billion and $7 billion for the plan, a source said. Isn't that a staggering number? In fact, this is equivalent to all properties in the whole of Spain combined, with the UK having a GDP of 314 trillion yuan, while the combined market capitalization of Microsoft and Apple, the largest publicly traded companies in the United States, is just over 6 trillion yuan. Looking at the integrated circuit industry, the total size of the global semiconductor integrated circuit market last year"Only"527 billion yuan, without Altman's project, it is expected to reach 1 trillion yuan by 2030 (McKinsey data).
Why did Ultraman invest so much money to build a new chip chain? The reason is simple, because it requires more computing power to support OpenAI's large-scale language models, such as GPT-5, etc. Altman tweeted that OpenAI is currently capable of creating about 100 billion words of content per day, while humans around the world can collectively create about 100 trillion words of content per day. This means that OpenAI's AI models have surpassed the ability of humans to produce words.
However, to keep up with the huge trend of model competition, it is necessary to build a strong computing power facility, and AI chips are becoming a bottleneck, limiting the development of OpenAI to a certain extent. Currently, OpenAI relies primarily on Nvidia's GPUs to run AI models, but Nvidia's GPUs are relatively limited and expensive. In the AI chip market, Nvidia controls about 80% of the share, and other competitors such as Google, Apple, Intel, AMD, etc. also have their own AI chip products, but it is difficult to compete with Nvidia.
Altman's chip initiative aims to disrupt this pattern and provide more options and advantages for OpenAI and other AI companies. Altman has reportedly held talks with representatives of the UAE**, SoftBank, chipmakers and even energy **. His goal is to build several integrated circuit factories around the world that will specialize in the production of integrated circuits for artificial intelligence, such as NPUs (neural network processors), which will replace GPUs. Altman is also considering an alliance with South Korean giant Samsung to leverage its technology and experience in integrated circuit manufacturing.
Altman's flat plate manufacturing plan is undoubtedly an ambitious and demanding project, but it is also a project full of risks and uncertainties. In fact, wafer fabrication is a high-investment, high-barrier-to-ceiling, and highly complex industry that requires a large number of professionals, equipment, materials, technology, and political support. In addition, Altman's project may also raise some geopolitical sensitivities, such as concerns about the security and reliability of the wafer chain in the United States, and considerations for the competitiveness and influence of AI technology in other countries.
However, if Ultraman's chip plan is successful, OpenAI can further expand its current leading edge in the field of artificial intelligence technology - after all, this industrial chain must be a combination of software and hardware, and give priority to supplying itself. And for other companies and AI users, there may also be more chip options and lower costs, thus promoting the wider application and development of AI technology.
Will Ultraman's chip plan really come to fruition soon? Will he be able to realize his great dream of building the world's largest AI chip empire and challenging Nvidia's supremacy? Let's wait and see.