OpenAI Is Making Its Own AI Chips to Challenge NVIDIA

According to The Financial Times, OpenAI is preparing to enter mass production of its in-house AI chips starting in 2026, in partnership with Broadcom, to meet soaring computational demands and reduce reliance on NVIDIA GPUs. Broadcom CEO Hock Tan revealed a chip design order worth as much as $10 billion, with insiders confirming the client as OpenAI; the customised chips are intended solely for internal use and are not planned for external sale.

As early as 2023, OpenAI’s CEO Sam Altman lamented GPU shortages that constrained the speed and reliability of its API services, and reports at the time suggested the company was already exploring building its own chips in talks with Broadcom and TSMC. Following the GPT-5 launch, OpenAI reportedly aimed to double its server capacity within five months and began pursuing in-house chip development to offset potential GPU shortages and cut long-term hardware costs.

NVIDIA remains dominant in the market, with a 56% year‑over‑year revenue rise in its latest results, while export restrictions to China have delayed shipments of its H20 chips. As production nears in 2026, the initiative is framed as strengthening OpenAI’s model and service stability and pushing a broader shift toward software–hardware co-design in AI infrastructure.