OpenAI, the prominent artificial intelligence company, has been engaged in discussions about developing its own AI chips since last year. These chips are crucial for the advancements we see in modern AI technology. Neural networks, the foundational algorithmic framework behind AI progress, heavily rely on specialized chips. Currently, most AI chips are manufactured by Nvidia, which has held a dominant position in the market for over a decade. However, OpenAI is now considering diversifying its chip suppliers and potentially collaborating with other chipmakers.
Nvidia has become a key player in the AI development ecosystem, providing not only chips but also software and services. Its chips are capable of performing complex AI tasks like image and face recognition, speech recognition, and text generation for chatbots such as OpenAI’s ChatGPT. In fact, Nvidia accounts for over 70% of AI chip sales today, with a significant market share in training generative AI models. For instance, OpenAI’s ChatGPT utilizes a Microsoft supercomputer powered by 10,000 Nvidia GPUs.
However, OpenAI is beginning to question its reliance on Nvidia, especially considering the escalating costs associated with running ChatGPT. According to an analysis by Bernstein analyst Stacy Rasgon, each ChatGPT query costs the company approximately 4 cents, and with increasing usage, these costs will continue to rise. OpenAI spends a staggering $700,000 per day to operate ChatGPT. Therefore, the company is contemplating the development of its own AI chips as a potential solution to reduce dependency on Nvidia and mitigate mounting expenses.
OpenAI’s CEO, Sam Altman, has emphasized two main concerns driving the push for AI chip development: the lack of cutting-edge processors to power their software and the escalating costs associated with running the necessary hardware infrastructure. While discussions on this matter are still ongoing within the company, recent reports indicate that Microsoft is also exploring the development of its own AI chips under the project name “Athena.” It remains uncertain whether OpenAI is involved in the same project, but if successful, Microsoft plans to make its AI chips more widely available as early as this year.
Moreover, rumors suggest that OpenAI has been evaluating potential acquisition targets to expedite its chip development efforts. Acquiring chip companies has proven successful for other tech giants like Amazon, which acquired Annapurna Labs in 2015 to bolster its chip capabilities. While OpenAI’s plans for custom chips, either through acquisitions or independent development, are still in progress, it is clear that such endeavors will require significant investments and may take several years to materialize.
Ultimately, regardless of the path OpenAI chooses, it will likely remain dependent on Nvidia’s chips for the foreseeable future. However, the company is actively exploring strategies to diversify its suppliers and reduce reliance on a single chip manufacturer. These efforts reflect OpenAI’s commitment to advancing AI technology while simultaneously addressing the growing operational costs associated with its popular AI models.
By Dashveenjit Kaur