OpenAI moves towards hardware independence: possible proprietary chips by 2026

A OpenAIcompany behind the famous ChatGPTis charting a bold path in the world of chips artificial intelligence. In a strategic move, the company is collaborating with industry giants like Broadcom and TSMC to develop its first custom chipwhile also expanding partnerships with AMD e Nvidia to meet the growing demand for AI infrastructure.

Also read
OpenAI launches ChatGPT app for Windows
Elon Musk throws a party to criticize OpenAI and try to recruit employees for his AI

Initially, OpenAI considered an entirely in-house approach, including the idea of ​​building its own network of chip factories. However, due to the high costs and time required, the company chose to focus on chip design, leaving manufacturing to specialized partners.

This strategy puts OpenAI on the same level as tech giants like Amazon, Meta, Google e Microsoftwhich seek to balance internal development and external partnerships to guarantee chip supply and manage costs. OpenAI’s decision to diversify its chip sources, including developing a custom chip, could have significant implications for the entire technology sector.

OpenAI’s first chip, developed in partnership with Broadcom, will focus on inference – the process of applying trained AI to new data. While there is currently greater demand for training chips, experts predict that the need for inference chips may surpass it in the future.

For this project, OpenAI assembled an elite team of around 20 professionals, including experienced engineers from Google. The chip is scheduled to be manufactured in 2026, through TSMC, the world leader in semiconductors.

Meanwhile, OpenAI is not abandoning its existing partnerships. The company remains one of Nvidia’s biggest buyers of GPUs and plans to utilize AMD’s new MI300X chips via Microsoft’s Azure platform.

This diversification is important for OpenAI, considering the high operational costs associated with training AI models and running services like ChatGPT. With a projected loss of $5 billion this year, against revenue of $3.7 billion, the company seeks to optimize its resources and reduce expenses.

OpenAI’s strategy demonstrates a balanced approach: innovating internally while maintaining strategic partnerships. This not only ensures access to cutting-edge technology like Nvidia’s Blackwell chips, but also positions the company competitively in the rapidly evolving landscape of artificial intelligence.

Source: Reuters