🟦OpenAI collaborates with Broadcom and TSMC to develop semiconductors for AI?

OpenAI, which handles the conversational AI 'ChatGPT', seems to be partnering with Broadcom and TSMC to design and manufacture semiconductors for AI.
reuters.com

🟦OpenAI collaborates with Broadcom and TSMC to develop semiconductors for AI

OpenAI has outsourced the production of Broadcom-designed semiconductors for AI to TSMC and is moving toward mass production in 2026. Conventional AI chips focused on training performance, but in the future, an efficient inference process will be essential, and OpenAI is developing new chips to respond to this trend.

OpenAI had previously considered launching a contract manufacturing business, but decided to focus on in-house designed chips in terms of technology management and operating costs.

🟦Why OpenAI is partnering with Broadcom and TSMC

This is to ensure a stable supply of AI chips and improve cost efficiency. In particular, training AI models and inference processing require enormous computational resources, which account for the majority of operational costs. Until now, NVIDIA’s chips have been at the center of the AI market, but due to high costs and unstable supply, OpenAI is trying to reduce costs with chips designed in-house.

In addition, the chip development team includes experienced engineers who were involved in the development of Google’s TPU (Tensor Processing Unit), and there are high expectations on the technical side.

🟦Summary

The partnership between OpenAI, Broadcom, and TSMC will greatly contribute to the development of AI technology and the diversification of the AI chip market. It enables efficient inference processing for AI applications, and is expected to further evolve in the AI field.

It is hoped that this partnership between Broadcom and TSMC will not only build a stable supply system for inference chips, but also lead to a reduction in power consumption in data centers.

Copied title and URL