OpenAI to Develop First AI Chip with Broadcom by 2026

• OpenAI is collaborating with Broadcom to produce its first in-house AI chip, aiming to reduce reliance on Nvidia and meet growing infrastructure needs.

Strategic Shift Toward Custom Silicon

OpenAI is reportedly preparing to launch its first proprietary artificial intelligence chip in 2026, according to sources cited by the Financial Times. The initiative is being developed in partnership with Broadcom, a major U.S. semiconductor manufacturer. Rather than offering the chip to external clients, OpenAI intends to deploy it internally to support its own AI infrastructure. Neither OpenAI nor Broadcom has publicly commented on the report as of this writing.

The move reflects OpenAI’s growing need for scalable computing resources to train and operate its generative AI models. These systems, including ChatGPT, require significant processing power, which has traditionally been supplied by third-party hardware. In recent years, OpenAI has relied on chips from Nvidia and AMD, but increasing demand has prompted the company to explore alternatives. Developing its own silicon could offer greater control over performance and cost.

Broadcom’s Expanding AI Portfolio

Broadcom CEO Hock Tan recently stated that the company anticipates notable growth in AI-related revenue by fiscal year 2026. During an earnings call, Tan revealed that a new customer had placed over $10 billion in infrastructure orders, though he did not identify the client. The timing and scale of the order suggest a possible link to OpenAI’s chip development plans. Earlier this year, Tan mentioned that four new clients were working closely with Broadcom on custom chip solutions.

This collaboration aligns with a broader industry trend, as major tech firms increasingly pursue bespoke hardware to optimize AI workloads. Companies like Google, Amazon, and Meta have already invested in developing their own chips tailored to machine learning tasks. Custom silicon allows for better integration with proprietary software and can improve efficiency across data centers. Broadcom’s role as a design and manufacturing partner positions it to benefit from this shift.

Fabrication and Supply Chain Considerations

Sources previously told Reuters that OpenAI’s chip design would be finalized within months and sent to Taiwan Semiconductor Manufacturing Company (TSMC) for fabrication. TSMC is a leading global foundry known for producing advanced semiconductors for clients such as Apple and Nvidia. By working with TSMC, OpenAI aims to ensure high-quality production while maintaining flexibility in its supply chain. The company has also continued to use AMD and Nvidia chips to meet short-term infrastructure demands.

Efforts to diversify chip sourcing reflect a strategic response to market volatility and rising costs. OpenAI’s internal chip project is part of a broader plan to reduce dependence on Nvidia, whose GPUs dominate the AI hardware space. The initiative could also help OpenAI manage long-term scalability as its models grow more complex. If successful, the chip may serve as a foundation for future hardware tailored to generative AI.