Charlie Kawwas, president of Broadcom's Semiconductor Solutions Group, said on Monday that OpenAI is not the mystery client valued at $10 billion announced during Broadcom's September earnings call.
Kawwas appeared on CNBC's "Squawk on the Street" alongside OpenAI President Greg Brockman to discuss their plans to jointly build and deploy 10 gigawatts of custom AI accelerators.
The deal was largely expected after analysts quickly pointed to OpenAI as a potential new $10 billion partner for Broadcom. But after the two companies officially announced their plans on Monday, Kawwas said OpenAI didn't fit that description.
"I would love to get a $10 billion [purchase order] from my good friend Greg," Kawwas said. "He hasn't given me that purchase order yet."
Broadcom did not immediately respond to CNBC's request for further comment.
OpenAI has been aggressively pursuing AI infrastructure deals, hoping to expand its computing power to meet anticipated demand. The startup, valued at $500 billion, has signed multi-billion dollar agreements in recent weeks with companies like AMD, Nvidia, and CoreWeave. Broadcom doesn't disclose its major networking customers, but analysts cite Google, Meta, and TikTok's parent company, ByteDance, as its top three. Broadcom CEO Hock Tan said on a quarterly conference call with analysts in September that a fourth-largest customer had placed a $10 billion order for custom AI chips.
Tan said on the call that the order boosted Broadcom's forecast for AI revenue next year, when AI products will begin shipping.
OpenAI and Broadcom, which have been collaborating for 18 months, said Monday that they will begin deploying custom chip racks by the end of next year. The project is expected to be completed in 2029.
"By building our own chips, we can embed the knowledge gained from creating cutting-edge models and products directly into the hardware, unlocking new capabilities and levels of intelligence," Brockman said in a press release.
OpenAI and Broadcom reach multi-billion dollar chip development deal
OpenAI and Broadcom are working together to develop and deploy 10 gigawatts of custom AI chips and computing systems over the next four years, a high-profile partnership aimed at meeting some of the startup's massive computing needs.
OpenAI plans to design its own graphics processing units (GPUs), which will enable it to incorporate its experience developing powerful AI models into the hardware powering future systems. Under the agreement announced Monday, these chips will be jointly developed by OpenAI and Broadcom, with Broadcom deploying them starting in the second half of next year.
People familiar with the matter said the new agreement is worth billions of dollars. The companies have not disclosed financial terms.
Broadcom's stock price rose nearly 10% in early trading.
Broadcom specializes in designing custom AI chips specifically for specific AI applications. Broadcom began collaborating with OpenAI on custom chips 18 months ago and has expanded the collaboration to include related components such as server racks and networking equipment.
The two companies said the new racks will rely on Broadcom's Ethernet and other connectivity technologies. People familiar with the matter said the racks will be deployed in data centers owned by OpenAI as well as those operated by third parties. This massive deal brings the total computing power OpenAI has agreed to purchase from chip giants Broadcom and Nvidia to 26 gigawatts, enough to meet more than twice New York City's summer electricity needs.
The startup will have to spend hundreds of billions of dollars to fund these deals. It expects to generate $13 billion in revenue this year, meaning it will have to increase sales exponentially to cover its computing expenses.
OpenAI, valued at $500 billion, is the world's most valuable startup and has launched some of the most popular AI tools, including ChatGPT and video generation software Sora. It has over 800 million weekly active users.
Broadcom helps tech companies produce custom chips that accelerate AI computing, which has fueled its growth. The company's stock has risen more than 80% in the past six months, pushing its market capitalization to over $1.5 trillion.
OpenAI CEO Sam Altman and its president and infrastructure executive, Greg Brockman, said they don't have enough computing power. Demand for AI products is growing rapidly, and the two said they hope to build new, large-scale data centers around the world to keep up. Over the past few months, OpenAI's deals with chip and cloud computing companies have fueled a global rally in tech stocks. Each agreement has further boosted expectations set by Altman, who has described the seemingly endless amount of computing needed to fuel the AI revolution. The deals have also surprised some competitors, whose forecasts for computing costs have been much more conservative.
According to people familiar with the matter, Altman recently told employees that OpenAI wants to build 250 gigawatts of new computing capacity by 2033, a plan that would cost more than $10 trillion by today's standards. He said OpenAI must create new financing vehicles to fund this massive buildout, but he offered few specifics.
Technology website Sources earlier reported on Altman's computing goals.
OpenAI's partners are betting big that Altman's vision of AI reshaping the world can be realized, and that the startup will be a major winner in this technological transformation. In September, consultants at Bain & Company estimated that a wave of AI infrastructure spending would require $2 trillion in annual AI revenue by 2030, exceeding the combined 2024 revenues of Amazon.com, Apple, Alphabet, Microsoft, Meta Platforms, and Nvidia.
Source: Content compiled from CNBC
