OpenAI Partners with Broadcom to Produce Its Own AI chips

OpenAI is producing its own AI chips. According to Reuters, U.S. chipmaker Broadcom is supporting the AI startup to design and produce its first AI chip. The Broadcom OpenAI chip deal will see the initial batch of chips shipped starting 2026.

OpenAI’s Move to Diversify Chip Supply

Semiconductors are critical in the development of AI by major tech companies such as Microsoft, Google, and Meta Platforms. OpenAI relies heavily on substantial computing power to train its AI models and run its systems.

In 2024, the AI startup started working with TSMC and Broadcom to design its first-inhouse chip to power its AI systems. OpenAI also started examining different options to reduce costs and diversify chip supply. In February, 2025, the AI startup reportedly started working on a plan to lessen its reliance on NVIDIA for chip supply by designing an in-house AI silicon.

In October 2024, Bloomberg reported that Broadcom and OpenAI were developing an inference chip design to run AI services after training. The AI startup plans to use the OpenAI AI chip inhouse in 2026 first before making it available to external customers. This move will be significant in reducing OpenAI’s Nvidia chip dependency.

On September 4, 2025, Broadcom CEO Mr. Tan Hock Eng referenced the partnership with OpenAI when he said the company had secured orders worth more than $10 billion in orders from a new client.

“Last quarter, one of these prospects released production orders to Broadcom,” Tan said.

Broadcom’s Revenue Growth

Broadcom is one of the chip makers that has benefited from the AI boom that has seen companies and AI startups spend billions of dollars in training new models and establishing data centers. The chip maker has moved to allay fears of slowing growth, informing investors that its outlook will improve significantly in the 2026 financial year.

In June 2025, Broadcom beat Q1 revenue estimates. The company reported revenue of $15 billion in the second quarter, surpassing analyst projections of $14.99 billion. On September 4, 2025, Tan said in 2026, the company expects AI revenue to grow by margins similar to the current year, which is between 50% to 60%. He added that with a new Broadcom OpenAI chip deal, the company’s revenue growth would accelerate substantially because the AI startup has “immediate and pretty substantial demand.”

Broadcom’s revenue mix differs from that of NVIDIA. NVIDIA’s GPU business leads in AI accelerator (GPU) sales, while Broadcom’s strengths include custom ASICs, networking silicon and infrastructure products. However, companies that run AI labs and data centers rely on Broadcom’s custom-designed semiconductors and networking equipment to handle their workloads.

Broadcom’s Market Performance

On September 5, 2025, Broadcom shares soared by nearly 9% in premarket trading following the $10 billion AI chip order from OpenAI. Broadcom stock has gained 32% in 2025. Shares of the chip maker have more than doubled since April when they were at their lowest. Over this period, the company has added about $730 billion to its market capitalization.

In recent days, share prices of AI-focused chipmakers have fluctuated. On August 29, 2025, shares of Marvell Technology, which competes with Broadcom closely in custom chip productions, dropped by 19% after the company missed its data center revenue targets. With Tan’s leadership, Broadcom has upgraded its networking equipment to transfer information better between AI chips in data centers.

The company is also working to attract clients who want to develop custom-designed chips for AI tasks. Following years of acquisitions, Broadcom has become a sprawling hardware and software giant. In June 2025, the company launched the Tomahawk Ultra Ethernet switch and started shipping it. The new switch is designed to improve AI workloads and support high-performance computing. Besides its AI work, the tech giant produces connectivity components for Apple’s iPhones and offers virtualization software.