OpenAI, the U.S.-based AI unicorn behind ChatGPT, is set to release its first self-designed AI chip next year in partnership with Broadcom, according to multiple sources.
The chip, referred to internally as an XPU, will be used exclusively by OpenAI to scale its AI infrastructure, rather than being sold to external customers.
A $10 Billion Commitment Drives Growth #
During a recent earnings call, Broadcom CEO Hock Tan revealed that a major new customer had committed to a $10 billion order for custom AI chips. While Broadcom did not officially disclose the client, sources confirmed the customer is OpenAI.
This makes OpenAI the fourth major client in Broadcom’s custom AI chip program, alongside the leading cloud providers. Tan noted that the deal has already created “immediate and significant demand,” prompting Broadcom to raise its AI revenue forecast for next year. Shipments of the new chips are expected to begin next year.
Technical Specs and Manufacturing #
OpenAI’s custom AI chip is rumored to leverage TSMC’s 3nm manufacturing process, integrating a systolic array architecture along with High Bandwidth Memory (HBM).
These design choices aim to maximize computational efficiency and memory bandwidth, critical for handling large AI models like GPT-5. OpenAI’s CEO Sam Altman emphasized that increasing computing capacity is essential to meet the growing demand for AI services and to accelerate model training. He recently stated that OpenAI plans to double its computing capacity within the next five months.
The Rise of Custom AI Chips #
Custom AI chips are becoming a dominant force in AI infrastructure. Tech giants including Google, Amazon, Microsoft, and Meta have previously developed their own chips to optimize AI workloads. Google’s TPU was also created in collaboration with Broadcom.
OpenAI’s move into in-house chip design reflects the broader trend of AI companies controlling more of their hardware stack to improve performance, reduce latency, and scale efficiently.
Market Impact #
The announcement has already positively affected Broadcom’s stock, which has surged more than 30% this year, with a 4.5% gain in after-hours trading following the earnings call.
By producing its own chips, OpenAI positions itself to accelerate AI model deployment, optimize costs, and maintain a competitive edge in the rapidly expanding AI market.
Conclusion #
OpenAI’s first self-designed AI chip, co-developed with Broadcom, marks a strategic leap in AI infrastructure, providing the computing power necessary for next-generation models like GPT-5. As demand for AI services continues to grow, custom AI hardware will play an increasingly critical role in supporting high-performance, scalable, and efficient AI workloads.