OpenAI links up with Broadcom to produce its own AI chips (arstechnica.com)

OpenAI is developing its own AI chip in partnership with semiconductor giant Broadcom, aiming to ship the custom hardware as early as next year. This marks OpenAI’s first foray into in-house chip design, reflecting a strategic push to meet the soaring demand for AI compute power and reduce dependence on Nvidia, the current dominant supplier of AI accelerators. The collaboration follows a trend among major tech companies like Google, Amazon, and Meta, who have similarly invested in bespoke silicon tailored to the intensive workloads of AI training and inference. The chip will be used exclusively by OpenAI for its internal operations rather than being sold externally, underscoring the company’s focus on optimizing its AI infrastructure at scale. Broadcom’s CEO Hock Tan hinted at a new major customer committing to $10 billion in orders during an earnings call, which sources confirmed is OpenAI. This partnership signals a significant shift in the AI hardware landscape, suggesting OpenAI’s readiness to take greater control over its performance and cost structure amid an escalating global competition for advanced AI chips. The move could accelerate innovation in AI model efficiency and deployment as the sector increasingly demands specialized silicon tailored to next-generation AI workloads.