OpenAI is planning to collaborate with the American chip giant (Broadcom) to design its own proprietary AI chips, aiming for a launch in 2026. These chips will be prioritized for OpenAI's internal computations.
OpenAI creates custom chips XPU
According to a report by Bloomberg, OpenAI is actively preparing a chip project, aiming for an official shipment in 2026. The new chip is internally codenamed "XPU" and is positioned to differentiate itself from GPUs produced by Nvidia and AMD.
These chips will initially be dedicated to OpenAI's own products and services, and only when testing is mature will they gradually open up for larger-scale applications. This strategy is similar to those of tech giants like Google, Amazon, and Meta, all aimed at ensuring that AI infrastructure can keep up with rapidly growing demand.
Broadcom is actively securing orders, revealing a mysterious billion-dollar customer.
Broadcom CEO Chen Fuyang revealed on September 4 that the company is currently assisting multiple clients in designing AI accelerators, emphasizing an optimistic market outlook. He even stated directly:
"A mysterious new client has promised orders exceeding $10 billion, which will allow Broadcom's AI revenue performance for the fiscal year 2026 to surpass original expectations."
Although Broadcom did not disclose the name of the client, it has been confirmed by the Financial Times and informed sources that this client is OpenAI.
OpenAI is the fourth largest AI chip customer of Cheng Bo Tong.
Broadcom announced in its latest financial report that it has successfully secured the custom AI chip business of its fourth large client, delivering results that exceeded Wall Street expectations.
Although the officials have not responded directly, it is widely believed that this order comes from OpenAI, indicating that the collaboration between the two parties has officially entered the mass production stage.
XPU vs GPU, the market landscape will undergo a transformation
OpenAI's XPU is viewed as the core computing foundation for the next-generation large language model GPT-5. Reports indicate that the relevant chips are expected to begin trial production as early as 2025, with a full rollout in 2026. If the plan goes smoothly, it will effectively reduce OpenAI's reliance on NVIDIA GPUs.
Currently, the demand for AI computing is mainly monopolized by Nvidia, and major companies have to queue up to purchase GPUs. OpenAI's entry into the self-developed chip field means that there will be an additional option in the market, which also opens up a new positioning for Broadcom.
( The market value once reached 1.15 trillion dollars! How did AI chip giant Broadcom AVGO break into Google, Meta, and AWS )
This article discusses OpenAI partnering with Broadcom to promote its own AI chips, which are expected to be released as early as 2026. It first appeared in Chain News ABMedia.
This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.
OpenAI teams up with Broadcom to launch its own AI chip, expected to be released as early as 2026.
OpenAI is planning to collaborate with the American chip giant (Broadcom) to design its own proprietary AI chips, aiming for a launch in 2026. These chips will be prioritized for OpenAI's internal computations.
OpenAI creates custom chips XPU
According to a report by Bloomberg, OpenAI is actively preparing a chip project, aiming for an official shipment in 2026. The new chip is internally codenamed "XPU" and is positioned to differentiate itself from GPUs produced by Nvidia and AMD.
These chips will initially be dedicated to OpenAI's own products and services, and only when testing is mature will they gradually open up for larger-scale applications. This strategy is similar to those of tech giants like Google, Amazon, and Meta, all aimed at ensuring that AI infrastructure can keep up with rapidly growing demand.
Broadcom is actively securing orders, revealing a mysterious billion-dollar customer.
Broadcom CEO Chen Fuyang revealed on September 4 that the company is currently assisting multiple clients in designing AI accelerators, emphasizing an optimistic market outlook. He even stated directly:
"A mysterious new client has promised orders exceeding $10 billion, which will allow Broadcom's AI revenue performance for the fiscal year 2026 to surpass original expectations."
Although Broadcom did not disclose the name of the client, it has been confirmed by the Financial Times and informed sources that this client is OpenAI.
OpenAI is the fourth largest AI chip customer of Cheng Bo Tong.
Broadcom announced in its latest financial report that it has successfully secured the custom AI chip business of its fourth large client, delivering results that exceeded Wall Street expectations.
Although the officials have not responded directly, it is widely believed that this order comes from OpenAI, indicating that the collaboration between the two parties has officially entered the mass production stage.
XPU vs GPU, the market landscape will undergo a transformation
OpenAI's XPU is viewed as the core computing foundation for the next-generation large language model GPT-5. Reports indicate that the relevant chips are expected to begin trial production as early as 2025, with a full rollout in 2026. If the plan goes smoothly, it will effectively reduce OpenAI's reliance on NVIDIA GPUs.
Currently, the demand for AI computing is mainly monopolized by Nvidia, and major companies have to queue up to purchase GPUs. OpenAI's entry into the self-developed chip field means that there will be an additional option in the market, which also opens up a new positioning for Broadcom.
( The market value once reached 1.15 trillion dollars! How did AI chip giant Broadcom AVGO break into Google, Meta, and AWS )
This article discusses OpenAI partnering with Broadcom to promote its own AI chips, which are expected to be released as early as 2026. It first appeared in Chain News ABMedia.