There was news that OpenAI had joined hands with Broadcom and TSMC to develop its own AI-only chips
News has surfaced that OpenAI has collaborated with Broadcom and TSMC to work on the development of its exclusive AI-only chips. This move marks a significant shift for OpenAI, which has traditionally heavily relied on Nvidia's GPUs. By diversifying the chip supply chain and aiming to lower costs, OpenAI is setting its sights on launching its own chip by 2026, as reported by Reuters.
Partnership Details
Broadcom, a renowned global semiconductor and communication equipment company, will be contributing to the design of the AI chips, while TSMC will oversee the production process. Initially, OpenAI had considered establishing its semiconductor factory for chip production. During this time, Sam Altman, a key figure at OpenAI, explored potential collaborations with Korean companies such as Samsung Electronics and SK. Ultimately, the decision was made to focus on chip design rather than venturing into production due to cost and time constraints.
Insights into Broadcom
Broadcom, established in 1991 by Henry Samueli and Henry Nicholas, is a key player in the semiconductor industry. The company, which originated in San Jose, California, specializes in designing semiconductor chips used in various applications such as network equipment, smartphones, storage devices, and communication equipment. In 2015, Avago Technologies acquired and merged with Broadcom Corporation, subsequently renaming the entity to Broadcom Limited.
Broadcom's expertise lies in ASIC (Application-Specific Integrated Circuit) and SoC (System-on-Chip) design, particularly excelling in customizing chips for specific applications. By integrating high-performance processors and memory into a single chip, Broadcom can efficiently execute complex operations, making it ideal for AI acceleration needs. The company also develops controllers and interface chips for storage, tailored to process and transfer significant volumes of data swiftly in data centers and cloud infrastructures.
Implications of the Partnership
OpenAI's collaboration with Broadcom carries considerable implications. Recent reports suggested that OpenAI is working on a cutting-edge AI model named "Orion," slated to be hosted in the Microsoft Azure data center. While access to Orion is currently limited, sources indicate that it will surpass the performance of the GPT-4 model by up to 100 times, positioning it as a significant advancement in AI capabilities.
Financial Considerations
To address the substantial computing demands, OpenAI is contemplating raising subscription fees to secure the necessary resources. With the company's estimated corporate value surpassing $157 billion, substantial investments are required to support large-scale language model development and operational sustainability.
