OpenAI is reportedly undertaking a collaboration with Broadcom to develop new custom silicon that is specifically designed to manage its substantial AI workloads required for inference, according to sources cited by Reuters. This initiative marks a significant step for OpenAI as it seeks to enhance its technological capabilities to meet the increasing demands of AI applications.
The company has established a specialised chip development team consisting of approximately 20 employees. Among these are leading engineers who previously contributed to the development of Google’s Tensor processors, well-known for their role in advancing AI-related functionalities. This move signifies OpenAI's dedication to building a bespoke solution aimed at optimising performance and efficiency for its AI operations.
Despite these advancements, production of the custom-designed hardware is not expected to commence until 2026, based on the current timeline. This projection reflects the complex nature of semiconductor development and the extensive timeline generally associated with the design, testing, and mass production phases.
Until these custom chips are in production, OpenAI is adapting its current technological infrastructure by incorporating AMD chips into its Microsoft Azure setup. AMD's MI300 chips, unveiled last year, have played a pivotal role in bolstering the company's presence in the competitive data centre market. The integration of AMD chips is seen as a strategic move to support OpenAI's immediate computing needs while awaiting the completion and implementation of its specialised hardware.
Previously, The Information reported OpenAI's discussions with Broadcom and other semiconductor designers regarding the potential creation of its own AI chip. Additionally, earlier in the year, Bloomberg highlighted OpenAI's intentions to establish its network of foundries, which are critical in the semiconductor manufacturing process. However, according to Reuters, these ambitious plans have been put on hold indefinitely, largely due to prohibitive costs and the lengthy timeframes involved.
Furthermore, OpenAI has secured manufacturing capacity with Taiwan Semiconductor Manufacturing Company (TSMC), a prominent player in the chip manufacturing industry. This arrangement likely aims to ensure that once its custom chips are ready for production, OpenAI will have the necessary facilities and expertise to produce them at scale.
Overall, these developments highlight OpenAI's continued efforts to innovate and tailor its infrastructure in response to the evolving needs of AI technology, while also underscoring the challenges and strategic decisions involved in advancing its hardware capabilities.
Source: Noah Wire Services