OpenAI Plans to Launch Its First In-House AI Chip

OpenAI plans to build its first in-house AI chip in 2026. For this purpose, the company partners with manufacturers Broadcom and TSMC (Taiwan Semiconductor Manufacturing Company), Reuters reports.

OpenAI has also added AMD chips alongside chips from Nvidia to support its AI systems and meet infrastructure demands.

With in-house developed products, the company aims to reduce costs, diversify its supply, and build a network of factories. Citing sources that preferred to stay anonymous, Reuters says for now OpenAI has dropped its plans to build foundries due to high costs and will focus on in-house chip design.

AI chips are needed to train systems like ChatGPT, which require substantial computing power. OpenAI now primarily relies on graphics processing units (GPUs) produced by Nvidia to train models on data and to make predictions or decisions based on the received information.

This process of AI training is expensive, factoring in hardware, electricity, and compute costs. In 2024, OpenAI has projected a $5 billion loss on $3.7 billion in revenue. So, OpenAI, Microsoft, Meta, and other development companies have been exploring in-house development and external alternatives.

Currently, Nvidia’s GPUs dominate the market, accounting for 80% of the share. However, rising competition and companies developing their own chips could change the situation.

Related: Only AI can restrain AI, claims Nvidia CEO Jensen Huang