- OpenAI contemplates crafting its artificial intelligence chips, evaluating potential acquisitions.
- CEO Sam Altman emphasizes the demand for more AI chips, citing the operational costs and chip shortages.
- OpenAI's decision on AI chip development might reshape its association with primary backer Microsoft.
October 6, 2023: OpenAI, the owner of ChatGPT, is considering the production of its artificial intelligence chips. As per Reuters, they have even looked into possible acquisition targets to facilitate this venture.
Although no concrete decision has been made, insiders reveal that OpenAI has been deliberating on ways to counteract the shortage of high-priced AI chips it depends on since the previous year.
The brainstormed solutions range from crafting their AI chip, enhancing collaboration with chipmakers such as NVIDIA, and expanding its supplier base beyond NVIDIA.
When approached, OpenAI chose to remain silent on the matter.
Sam Altman, the CEO of OpenAI, has voiced his concerns about the scarcity of graphic processing units (GPUs), a segment that sees NVIDIA holding a substantial market share, exceeding 80%.
Altman emphasizes the need for more AI chips, alluding to the challenges faced due to a lack of advanced processors that drive OpenAI’s software and the significant costs linked to operating the hardware essential for their projects and products.
For context, since 2020, OpenAI’s generative artificial intelligence technologies have been powered by an expansive supercomputer developed by Microsoft, one of OpenAI’s primary supporters. This supercomputer incorporates 10,000 of NVIDIA’s GPUs.
The operational expenses for running ChatGPT are notably high. As per an analysis by Bernstein’s Stacy Rasgon, each query roughly amounts to 4 cents.
If ChatGPT’s queries escalate to even a tenth of Google search’s magnitude, an initial investment of approximately $48.1 billion in GPUs would be essential. Subsequent annual expenses to maintain the operations would be around $16 billion in chips.
Venturing into AI chip development would align OpenAI with tech behemoths like Alphabet’s Google and Amazon.com, which have consistently sought autonomy over their chip designs.
The endeavor to construct a custom chip is daunting and demands considerable resources, potentially running into hundreds of millions annually. Furthermore, committing to this initiative doesn’t assure success.
OpenAI might expedite the custom chip creation process by acquiring a chip firm.
Amazon followed a similar path with its acquisition of Annapurna Labs in 2015.
OpenAI had progressed in this direction to a point where they had meticulously assessed a potential acquisition target. However, the identity of this prospective purchase remains undisclosed.
Should OpenAI decide to pursue the creation of a custom chip, including a possible acquisition, it’s a lengthy endeavor that will span years. This timeline means their reliance on market leaders like NVIDIA and Advanced Micro Devices will persist for the foreseeable future.
Other tech giants, including Meta, have experienced challenges in their custom chip ventures. According to a Reuters report, Meta is now crafting a newer, more comprehensive AI chip after confronting issues with its AI chips.
Interestingly, Microsoft, a primary supporter of OpenAI, is also devising a custom AI chip. OpenAI is testing this chip, hinting at an evolving dynamic between the two corporations.
With the advent of ChatGPT last year, there’s been a surge in demand for specialized AI chips. These unique chips, termed AI accelerators, are pivotal for training and implementing the latest generative AI technology.
NVIDIA is one of the few entities manufacturing these coveted AI chips, establishing a strong market presence.