The Bad News for Broadcom in OpenAI’s AMD Deal

Shares of Broadcom are up 46% this year, while AMD stock has climbed 73%. (Justin Sullivan/Getty Images)

Key Points

• Advanced Micro Devices secured a long-term deal to supply 6 gigawatts of Instinct GPUs to OpenAI, starting with 1 gigawatt in late 2026.
• AMD anticipates the OpenAI partnership will generate tens of billions in annual revenue and over $100 billion in total revenue over several years.
• The deal positions AMD as a key AI chip supplier, with its GPUs offering flexibility over Broadcom’s ASICs for evolving AI algorithms.

The battle for second place in the race for artificial-intelligence chip leadership just got more interesting.

Analysts have long wondered whether Broadcom would become the clear number two player in AI chips after Nvidia . But the size of Advanced Micro Devices’ latest deal with OpenAI is now an order of magnitude larger than the one it has with Broadcom, giving AMD a leg up in becoming the secondary AI chip supplier of choice.

AMD stock surged on Monday after the chip maker announced a long-term deal to become a key supplier to OpenAI’s artificial-intelligence infrastructure buildout.

OpenAI plans to deploy 6 gigawatts worth of AMD Instinct graphics processing units, or GPUs, over the course of the partnership, with the first 1 gigawatt deployment beginning in the second half of 2026. For context, one nuclear reactor can typically generate one gigawatt of electricity.

OpenAI is the leader in AI technology with market-leading products such as ChatGPT, AI models for enterprises, and the recently released social video app Sora. The suppliers that OpenAI chooses can also serve as a positive signal for the rest of the industry to follow.

AMD CEO Lisa Su said the deal is a “clear validation” of the company’s technology road map and that OpenAI will use AMD chips for both training and inference. Inference is the process of generating answers from AI models.

AMD management said they expect the OpenAI partnership will enable the company to eventually generate tens of billions of dollars in annual revenue and over $100 billion in total revenue from selling chips over the next several years.

While the figures are large, they are still much smaller than Nvidia’s numbers. According to Wall Street estimates, Nvidia’s data center revenue is expected to rise from $185 billion this fiscal year ending in January to $300 billion by fiscal 2029. The AMD agreement “is all incremental to our work with NVIDIA (and we plan to increase our NVIDIA purchasing over time),” OpenAI CEO Sam Altman said on social media Monday.

Last month, Broadcom announced it had won a $10 billion order for AI chips from a new customer, which The Wall Street Journal reported was OpenAI. Broadcom helps large technology companies design custom chips for AI with its high-end artificial intelligence application-specific integrated circuits, or ASICs. The AMD deal may speak to the advantages of GPUs over ASICs as AI technology continues to advance.

On an earnings call earlier this year, AMD’s Su explained why she believed GPUs would dominate the future and ASICs would get a smaller piece of the market. ASICs are “specific workload optimized, whereas GPUs will enable significant programmability and adjustments to all of these [AI] algorithm changes,” she said.

The basic message is that GPUs are flexible and a perfect fit for the current and future rapidly changing landscape in AI.

Broadcom has been put on watch.

Write to Tae Kim at tae.kim@barrons.com