OpenAI’s founder, Sam Altman, attended Intel’s IFS Direct Connect conference on February 20th, sparking speculation about his intentions in the semiconductor industry. Altman had recently been rumored to be raising $7 trillion to build his own chip factory, and his appearance at the conference only fueled speculation about his ambitions in the chip industry.
The idea of Altman building his own chip factory has been met with skepticism by industry experts. Semiconductor architect Jim Keller even stated that he could achieve the same results with less than $1 trillion, suggesting that Altman’s approach may not be necessary.
In terms of the amount of money involved, Altman’s proposed investment is astronomical. With global GDP totaling around $105 trillion per year, $7 trillion represents about 6% of that amount, making it even larger than the combined market value of tech giants like Apple. Additionally, Altman’s background in AI creates a significant gap between him and the chip industry. So why is Altman’s ambitious move not being well-received?
The main reason behind Altman’s interest in entering the chip industry is cost. As ChatGPT gained popularity worldwide by the end of 2022, big companies started actively developing their own AI language models and related applications. Microsoft, Amazon, and Google, among other cloud giants, have almost all launched their own language models by 2023, challenging ChatGPT’s dominance. According to Cheng Shijia, CEO of iCar, Altman realized that these companies were catching up, creating a different competitive landscape compared to a year ago.
If we extend the timeline a bit, it becomes clear that as AI productivity tools continue to develop, smaller-scale OpenAI will no longer be able to maintain its lead solely through models like ChatGPT. They will also struggle to compete with cloud giants in terms of chip orders and cost reduction. This is considered Altman’s biggest nightmare, as explained by Cheng Shijia. He further elaborated that large cloud providers not only have dedicated teams designing their own AI chips but have also invested heavily in infrastructure such as large-scale data centers. This advantage allows them to lower the cost of computing power, i.e., chip prices. On the other hand, OpenAI does not possess these conditions, which is why Altman is in a hurry, knowing that infrastructure will ultimately determine the winner.
However, achieving success in the chip industry may not be easy for Altman, despite his determination to compete with cloud giants. It is rumored that he has approached TSMC, Intel, and Samsung for potential collaborations. Cheng Shijia believes that there are three key points worth considering in this regard.
Firstly, AI is in a rapidly evolving phase, with significant differences in technological advancements within a year. Altman needs to consider that hardware will continue to progress, and AI software architecture will also change. This poses a risk, as there is still much room for optimizing AI chips. In this situation, investing in a chip factory without careful consideration may lead to irreversible consequences and increase the difficulty of fundraising.
Secondly, geopolitics plays a crucial role in the semiconductor industry, as it has become a strategic asset for many countries. Building a chip factory involves various geopolitical considerations. As TSMC is a Taiwanese company, replicating its success in the United States remains uncertain. Furthermore, complex political factors come into play, making it a topic of discussion where the factory should be located and when it should be established, even if Altman manages to raise $7 trillion.
Lastly, there are reports that NVIDIA CEO Jensen Huang is preparing to establish a custom chip (ASIC) department. This is essentially a signal to AI giants that they should not build their own chips. Guo Dajing, General Manager of the Chinese Development and Innovation Accelerator, believes that the semiconductor industry is highly specialized, and building an in-house IC design team is not the most suitable option. Cheng Shijia also believes that based on the available information, the feasibility of Altman’s plan seems low.
In conclusion, while Altman’s determination to compete with cloud giants is commendable, realizing his goals may prove to be challenging. The semiconductor industry is complex and rapidly evolving, and Altman must consider the risks and uncertainties associated with building a chip factory. The geopolitical landscape and the emergence of alternatives, such as NVIDIA’s custom chips, further complicate the situation. Therefore, it is understandable why some experts have reservations about Altman’s plans.