ChatGPT creator, OpenAI, has begun exploring ways of making its own artificial intelligence chips, going as far as evaluating a potential acquisition target.
According to recent internal discussions, the OpenAI has not yet decided to move ahead with its plan, however, it has discussed various options to solve the shortage of expensive AI chips that it relies on.
Sources from OpenAI say the options discussed include the company working more closely with other AI chip developers including Nvidia to build its own AI chip, and also diversifying its suppliers beyond Nvidia.
Demand for specialized AI chips has soared since the launch of ChatGPT last year. Specific chips, or AI accelerators, are necessary to train and run the latest generative AI technology. Nvidia, which dominates the graphics processing units (GPU) market, is one of the few chip developers that produces powerful AI chips.
OpenAI CEO, Sam Altman, who has already complained about the Nvidia-dominated GPU market, has moved the acquisition of more AI chips to the firm’s top priority list. Nvidia, currently controls more than 80% of the global market for the chips that can reliably run AI applications.
Altman has identified two major concerns which have propelled OpenAI’s efforts to get more chips. These being, a shortage of the advanced processors that power OpenAI’s software and the high costs associated with running the hardware necessary to power its efforts and products.
Since 2020, OpenAI has developed its generative AI technologies on a massive supercomputer constructed by Microsoft, who is among its largest backers. The supercomputer uses 10,000 of Nvidia’s GPUs.
Keeping ChatGPT running is very expensive for the company. If the chatbot’s queries grow to a tenth the scale of Google search, it would require roughly $48.1 billion worth of GPUs initially and about $16 billion worth of chips a year to keep operational.
OpenAI developing its own AI chips would place it among tech players such as Google and Amazon.com, who are also working to design their own AI chips for their businesses.
Should OpenAI decide to build a custom chip, it would be a major strategic initiative and a heavy investment resulting in annual costs of hundreds of millions of dollars. Alternatively, the company could acquire an existing AI chip company instead, which could speed up the process of it building its own AI chips.
OpenAI seems to be leaning for the latter path, and according to Reuters, it has performed due diligence on a potential acquisition target. The identity of the examined company was not given, however.
Until OpenAI builds its own chip, which could likely take several years, the company will still be dependent on commercial providers like Nvidia and Advanced Micro Devices (AMD).
