OpenAI, the maker of ChatGPT, is considering developing its own artificial intelligence chips and has even considered analyzing a potential acquisition target.
However, since at least last year, it has addressed potential solutions to the problem of the pricey AI chip shortage on which OpenAI is dependent.
These choices include developing its own AI processor, collaborating more closely with other chipmakers, such as Nvidia, and expanding its supply base.
OpenAI chose not to respond.
The company’s primary aim, according to CEO Sam Altman, is to buy more AI chips. In a market dominated by Nvidia which holds more than 80% of the global market share for the chips, best suited to powering AI applications, he has openly lamented the lack of graphics processing units.
The drive to obtain more chips is related to two key issues that Altman has identified: a lack of the cutting-edge processors needed to power OpenAI’s software and the “eye-watering” costs of maintaining the hardware required to support its initiatives and products.
Since 2020, Microsoft, one of OpenAI’s biggest investors, has built a gigantic supercomputer using 10,000 Nvidia graphics processing units (GPUs), on which the company has been developing its generative artificial intelligence technology.
The cost of running ChatGPT is prohibitive for the business. According to an analysis by Bernstein analyst Stacy Rasgon, each query costs about 4 cents. Initially, ChatGPT would need about $48.1 billion in GPUs, and it would need about $16 billion in chips a year to run if searches reached a tenth the size of Google searches.
ERA OF CUSTOM CHIPS
OpenAI would join a select group of major tech companies, including Alphabet’s Google and Amazon.com, that have pushed to gain control over producing the chips that are essential to their operations.
It’s unclear if OpenAI will carry out its ambition to develop a unique chip. According to experts in the field, doing so would need a significant investment and strategic effort that may cost hundreds of millions of dollars annually. Even if OpenAI invested resources on the project, success was not guaranteed.
A chip business purchase might expedite the development of OpenAI’s own processor, as it did for Amazon.com when it acquired Annapurna Labs in 2015.
OpenAI had thought about the approach up until the point where it carried out due diligence on a prospective acquisition candidate.
It was impossible to determine the name of the corporation whose purchases OpenAI looked into.
Even if OpenAI moves forwards with its ambitions for a custom chip, which could include an acquisition, the endeavor is likely to take many years, making the business dependent on for-profit suppliers like Nvidia and Advanced Micro Devices in the interim.
With mixed results, some major tech corporations have spent years developing their own processors. Meta’s bespoke chip project has been plagued by problems, forcing the company to abandon several of its AI processors. The owner of Facebook is now developing a newer processor that will support all forms of AI activity.
Microsoft, the primary supporter of OpenAI, is also producing a unique AI chip that OpenAI is evaluating. The initiatives might represent a deeper rift between the two businesses.
Since the introduction of ChatGPT last year, the demand for specialised AI chips has skyrocketed. The most recent generative AI technologies must be trained and run on specific chips, or AI accelerators. One of the few semiconductor manufacturers that produces practical AI chips, Nvidia rules the market.
