OpenAI, the company renowned for its creation of ChatGPT, is actively considering the development of its own artificial intelligence chips, with internal discussions even touching upon the prospect of acquiring a chip manufacturing entity. Although no definitive decision has been made at this point, insiders familiar with the matter reveal that OpenAI has been grappling with ways to address the ongoing scarcity of costly AI chips that are vital to its operations.
Among the potential solutions discussed are crafting proprietary AI chips, forging deeper collaborations with chip giants like Nvidia, and diversifying chip suppliers beyond Nvidia’s dominance in the market. OpenAI, when contacted for comment, opted to remain silent on the matter.
CEO Sam Altman has underscored the acquisition of additional AI chips as a critical objective for the organization. He has publicly lamented the scarcity of graphics processing units (GPUs), an arena controlled predominantly by Nvidia, which commands over 80% of the global GPU market, particularly suitable for AI applications.
OpenAI’s quest for more chips is motivated by two primary concerns outlined by Altman: the dearth of advanced processors essential for the company’s software and the exorbitant costs associated with maintaining the requisite hardware to power its initiatives and products.
Since 2020, OpenAI has been utilizing a vast supercomputer, constructed by Microsoft, one of its principal backers, which employs 10,000 of Nvidia’s GPUs. Running ChatGPT incurs substantial expenses, estimated at around 4 cents per query, according to Bernstein analyst Stacy Rasgon. If ChatGPT queries were to reach a scale one-tenth that of Google search, it would necessitate approximately $48.1 billion worth of GPUs initially and about $16 billion worth of chips annually to sustain operations.
Pioneering its own AI chip development would position OpenAI among a select group of prominent tech players, including Google and Amazon, who have taken strides towards designing chips essential to their core operations. The feasibility of this endeavor remains uncertain and would necessitate a considerable financial investment, potentially in the hundreds of millions of dollars annually, according to industry experts. Even with the allocation of substantial resources, success is not guaranteed.
An acquisition of an existing chip company could expedite OpenAI’s pursuit of proprietary chips, akin to Amazon’s acquisition of Annapurna Labs in 2015, but the identity of the potential acquisition target remains undisclosed.
Regardless of whether OpenAI ultimately embarks on the journey of developing custom chips, the undertaking is anticipated to span several years, rendering the company reliant on commercial providers such as Nvidia and Advanced Micro Devices during this interim period.
Several tech giants have ventured into custom chip development over the years with mixed results, as seen with Meta’s struggles. OpenAI’s principal supporter, Microsoft, is also in the process of developing a custom AI chip for OpenAI, indicating potential shifts in their partnership.
The demand for specialized AI chips has surged since the launch of ChatGPT last year, as they are indispensable for training and running the latest generative AI technology. Nvidia, being one of the few manufacturers of these essential AI chips, currently commands a dominant position in the market.
