OpenAI prepares to produce AI chips in collaboration with Broadcom and TSMC | Cpu vs gpu vs ram | Gpu vs cpu vs Graphics Card | Computer Hardware Parts List | Turtles AI

OpenAI prepares to build its own AI chips, collaborating with Broadcom and TSMC

Highlights:

• OpenAI collaborates with Broadcom and TSMC to develop custom AI chips.
• Altman’s strategy shift, focusing on a joint venture instead of own factories.
• Availability of new chips is not expected before 2026.
• Broadcom and TSMC at the center of discussions for production capacity.

Microsoft-backed OpenAI is looking to develop its own AI chips to address supply shortages and reduce dependency on Nvidia. The collaboration with Broadcom and talks with TSMC are key to this strategy.

The growing demand for computational capacity for AI is pushing OpenAI to explore new horizons in the semiconductor sector. Recently, the company initiated talks with Broadcom and other chip designers to develop custom solutions for its AI products, aiming to overcome issues related to supply chain limitations. This strategic move aims to reduce reliance on Nvidia, the current leader in producing GPUs necessary for AI applications. Aware of the challenges involved in chip production, OpenAI has also hired former Google employees specializing in this field.

A Bloomberg report last January indicated that OpenAI CEO Sam Altman was seeking to raise billions of dollars to build a network of chip factories. However, after a series of discussions with Taiwanese giant TSMC, Altman changed his approach. Instead of starting his own production plants, he is now focusing on a joint venture with Broadcom for chip development. This decision was influenced by initial estimates that had raised concerns among investors: Altman had calculated a cost of about $7 trillion to meet the demand for chips required for AI applications, a figure deemed impractical.

According to sources from The Information, the new chips designed by OpenAI are expected to achieve performance similar to Nvidia’s products, but they are not expected to be available on the market until at least 2026. Chip design requires established expertise, and although the process has been simplified thanks to semiconductor manufacturers like TSMC, it remains an endeavor that demands time and dedication.

The talks between Altman and TSMC have significantly impacted OpenAI’s strategy. TSMC has expressed willingness to allocate production capacity for OpenAI, provided that the latter commits to significant orders. This partnership could represent a crucial step for OpenAI in the semiconductor sector, diverting some production capacity from Nvidia.

Meanwhile, Broadcom has been at the center of attention in the AI chip sector thanks to its networking and ASIC products, essential for AI server connectivity and customized solutions for developers. Broadcom’s shares have seen a 78% increase over the past twelve months, while Nvidia’s stock has risen by 157% in the same period.

According to another source cited by the Financial Times, OpenAI does not intend to "sit back and let others build" the energy, chip, and compute capacity required for AI workloads, especially "when they are on the front line." These plans require "massive sums of money" and highlight OpenAI’s commitment to directly addressing the challenges of AI hardware supply.