OpenAI is considering delving into the AI chip domain and has even scrutinized potential acquisition prospects, sources reveal.
While a final decision hasn’t been made, OpenAI has been deliberating solutions to the AI chip scarcity, a pivotal resource for them, as per insiders. Contemplated solutions include developing proprietary AI chips, collaborating closely with chip manufacturers like Nvidia, and extending their supplier base beyond just Nvidia. OpenAI refrained from commenting.
CEO Sam Altman views acquiring more AI chips as imperative. He’s publicly voiced concerns about the scarcity of graphics processing units (GPUs) predominantly controlled by Nvidia, which holds a significant 80% global market share for AI-optimized chips.
Altman’s urgency for more chips emanates from two major worries: the dearth of sophisticated processors for OpenAI’s software and the staggering operational costs of the required hardware.
OpenAI and Microsoft collaboration
Since 2020, OpenAI has built its AI technologies on a colossal supercomputer by Microsoft, one of its primary supporters, incorporating 10,000 Nvidia GPUs. Operating ChatGPT-4 entails substantial costs, amounting to about 4 cents per query as indicated by Bernstein’s Stacy Rasgon. The expenses would skyrocket to around $48.1 billion in GPUs if ChatGPT’s queries equaled a tenth of Google search’s scale, with an additional $16 billion annually for sustainability.
OpenAI’s potential move to craft its AI chips would align it with a select tech elite, like Google and Amazon, who are keen on customizing chips integral to their operations.
The feasibility of OpenAI launching its tailored chip remains ambiguous. Embarking on such a venture demands significant investment, possibly hundreds of millions annually, as per industry insiders. Resource allocation to this project doesn’t assure success either.
Acquiring a chip firm could expedite OpenAI’s chip-making endeavors, reminiscent of Amazon’s purchase of Annapurna Labs in 2015.
OpenAI has been so invested in this path that it’s conducted an in-depth review of a potential acquisition candidate, an insider disclosed. However, the company’s identity remains undisclosed.
If OpenAI pursues the custom chip route, including potential acquisitions, the endeavor would span years. This would entail reliance on commercial vendors like Nvidia and AMD in the interim.
While several tech giants have attempted to create their chips, results have been inconsistent. Meta faced challenges with its custom chip, leading to discontinuations of certain AI chips. The company is now working on a more encompassing AI chip.
Meanwhile, OpenAI’s primary supporter, Microsoft, is also venturing into custom AI chip development, which OpenAI is currently examining. This could indicate a potential divergence between the two entities.
Conclusion
The launch of ChatGPT has fueled the demand for specialized AI chips. These distinct chips, dubbed AI accelerators, are indispensable for pioneering generative AI tech. Nvidia remains a dominant player in producing these pivotal AI chips.
Creating custom AI chips for GPT-5 Model could result in a breakthrough in performance and efficiency, allowing for faster computations and reduced operational costs.
Read related articles: