IT Specialist Media The Information Reports
Intention to Reduce Dependence on Nvidia GPUs
Cost-Efficient Model 'GPT-4o Mini' Released
OpenAI, the developer of ChatGPT, is reportedly in talks with U.S. semiconductor design company Broadcom to develop its own artificial intelligence (AI) chips, according to IT media The Information on the 18th (local time).
According to sources familiar with the matter, OpenAI is discussing the development of its own AI chips with Broadcom to ensure a smooth supply of graphics processing units (GPUs) necessary for developing powerful AI models. Additionally, OpenAI is recruiting engineers who previously participated in in-house chip development at Google to also push forward AI server chip development.
The development of in-house AI chips is part of a global semiconductor infrastructure plan envisioned by Sam Altman, CEO of OpenAI. Bloomberg reported earlier this year that CEO Altman is raising billions of dollars to build a semiconductor chip manufacturing network, with potential partners including Intel, TSMC, and Samsung Electronics. If all plans come to fruition, OpenAI will have all the major semiconductor processes required for AI development, from fabless (semiconductor design) to foundry (semiconductor contract manufacturing).
OpenAI is not the only company pursuing in-house AI chip development. With the AI boom sparked by Nvidia and the skyrocketing prices of GPUs essential for AI development, big tech companies such as Apple, Google, and Amazon are accelerating their own AI chip development to reduce dependence on Nvidia.
Sources predict that even if the in-house AI chip development is completed, many detailed issues need to be resolved, so production is expected to be possible only by 2026. Broadcom is known to have a division that produces application-specific integrated circuits (ASICs) for specific purposes and supplies them to other companies. On this day, Broadcom’s stock price closed at $160.52, up 2.91% from the previous session, on the New York Stock Exchange.
Meanwhile, OpenAI announced plans to release a smaller version of its latest AI model, GPT-4o, called 'GPT-4o Mini.' The aim is to reduce cost burdens for startups by offering it at a price more than 60% cheaper than the previous version, GPT-3.5 Turbo. ChatGPT Plus and Team subscribers will have access to the Mini model starting today.
© The Asia Business Daily(www.asiae.co.kr). All rights reserved.
