OpenAI is preparing to begin large-scale production of its own artificial intelligence chips next year in a bid to secure the massive computing power needed to run its services and reduce dependence on NVIDIA, the Financial Times reported on Friday.

According to the report, the company has designed a custom processor in partnership with US semiconductor manufacturer Broadcom, whose CEO recently disclosed a new $10 billion order from an unnamed client. Sources confirmed the client to be OpenAI. Unlike other chipmakers, OpenAI reportedly has no plans to sell the processors externally and intends to use them exclusively to power its own infrastructure.

The move marks a major step in the company’s strategy to address GPU shortages that have strained global AI development. Back in 2023, Reuters revealed that OpenAI was exploring custom chip design after CEO Sam Altman publicly cited shortages as a cause of API slowdowns and service reliability issues. At the time, OpenAI was said to be collaborating with both Broadcom and Taiwan Semiconductor Manufacturing Co. (TSMC). It remains unclear whether TSMC is still involved in the current production plan.

OpenAI’s in-house processors, described as “XPUs”, could eventually secure a significant share of the AI hardware market, where NVIDIA currently dominates. The company’s GPUs remain the gold standard for training and deploying advanced AI systems. NVIDIA’s financials reflect that dominance: its second-quarter revenue ending July 27 jumped 56 percent year-on-year, despite restrictions preventing shipments of its latest H20 chips to China.

The custom chip initiative comes after the rollout of GPT-5, which has sharply increased computing demand across OpenAI’s platforms. Altman has said the company will double its compute fleet within five months, prioritising access for paid ChatGPT subscribers. Producing its own chips would not only guarantee supply but could also reduce long-term costs associated with buying GPUs at premium prices.

Industry analysts say if OpenAI’s XPU project succeeds, it could reshape the competitive landscape of AI hardware, challenging NVIDIA’s dominance and accelerating the adoption of specialised processors optimised for generative AI workloads.