OpenAI another heavyweight? Altman expresses excitement about the "upcoming chip"

In addition to the rumored new model GPT-5.2 that surpasses Google's Gemini 3, OpenAI may have another major release coming soon. On Thursday, December 11th, Eastern Time, OpenAI CEO Sam Altman seemingly let slip during an interview after reaching an agreement with Disney that they are excited about "upcoming chips," but did not disclose any details.

Altman stated on Thursday that the release of the new model will increase market share and mentioned that market demand is "quite extreme." He also noted that Disney's $1 billion investment in OpenAI and the licensing agreement for iconic cartoon characters will bring incremental revenue, with demand for Disney characters "exceeding expectations." Disney CEO Bob Iger stated to the media on the same day that this investment is "a way" to enter the artificial intelligence (AI) field and will have a significant long-term impact on Disney's business.

Altman expressed enthusiasm for the chip plans, but OpenAI has yet to make any formal announcements regarding the progress of its self-developed chips. Since the beginning of this year, there have been continuous reports about OpenAI's chip development and related partnerships, indicating that the company is actively advancing its strategy to reduce dependence on Nvidia.

For investors in the AI chip sector, the advancement of OpenAI's self-developed chips could reshape the AI infrastructure market landscape, creating opportunities for suppliers beyond Nvidia while intensifying competition among tech giants in the custom chip field.

Chip Plans Come to Light

Since the beginning of this year, multiple media outlets have reported on the progress of OpenAI's chip project. Wallstreetcn mentioned that in February, reports indicated that OpenAI planned to complete its first custom chip design this year and send it to TSMC for manufacturing, aiming for mass production by 2026. The chip development team is led by former Google engineer Richard Ho and currently consists of about 40 people, collaborating with Broadcom to design chips primarily for running AI models.

In September, media reports stated that OpenAI reached an agreement with Broadcom to jointly design autonomous AI chips, with plans to officially begin mass production next year. Broadcom CEO Hock Tan subsequently announced during an earnings call that the company had secured a major customer for its fourth custom AI chip business, which committed to a $10 billion order. Sources confirmed to the media that this customer is indeed OpenAI.

On October 13th, OpenAI signed a formal agreement with Broadcom to announce their collaboration on developing custom chips and network devices, planning to deploy 10 gigawatts (GW) of AI data center capacity. This scale is equivalent to the power generation of about five Hoover Dams. The two parties plan to start deploying server racks in the second half of 2026, with the entire hardware deployment expected to be completed by the end of 2029

Multi-party Cooperation Layout

In addition to Broadcom, OpenAI has also been reported to be collaborating with other chip manufacturers.

On October 13, the same day OpenAI announced its partnership with Broadcom, media reported that OpenAI is in discussions with Arm, a semiconductor design company under SoftBank, planning to use Arm-designed central processing units in jointly designed AI server chips. As one of OpenAI's largest shareholders, SoftBank holds nearly 90% of Arm and has invested billions of dollars in OpenAI.

This year, OpenAI has also signed large-scale agreements with traditional chip suppliers. In September, NVIDIA announced it would invest up to $100 billion in OpenAI to support new infrastructure development, aiming for at least 10GW capacity. In October, OpenAI announced an agreement with AMD to deploy 6GW of AMD processors over several years.

In September, media mentioned that a person close to the project stated that OpenAI plans to use self-developed chips internally rather than providing them to external customers. Unlike the "GPUs" designed by suppliers like NVIDIA and AMD, Broadcom's custom AI chips are referred to as "XPUs."

Computing Power Demand Driven

Altman has been vocal about the urgent need for computing power. In August, he stated that considering the growing demand from the latest model GPT-5, the company is prioritizing computing power issues and plans to "double the scale of its computing cluster within the next five months."

Through custom chips, OpenAI can directly embed the experience accumulated in AI model development into hardware. After announcing the agreement with Broadcom in October, Altman stated that this would "unlock new capabilities and levels of intelligence." However, OpenAI co-founder and president Greg Brockman admitted that even 10GW of computing power is insufficient to support the company's vision of achieving general artificial intelligence.

OpenAI is not the only tech company exploring ways to reduce dependence on NVIDIA. Currently, NVIDIA holds about 80% of the AI chip market, and companies like Microsoft and Meta are also developing custom chips. Meta plans to invest $60 billion in AI infrastructure over the next year, while Microsoft plans to invest $80 billion by 2025. These interconnected deals have also heightened market concerns about an AI spending bubble