Broadcom stock soars 10% on OpenAI custom chip deal, adding to Nvidia and AMD agreements

Broadcom's stock experienced a significant surge of over 10% in premarket trading following the announcement of a formal partnership with OpenAI. This collaboration aims to co-develop and deploy a staggering 10 gigawatts of custom artificial intelligence accelerators, marking a pivotal step in enhancing AI infrastructure on a global scale. The partnership, which has been in the works for the past 18 months, is set to result in the production of racks filled with OpenAI-designed chips, with deployment expected to commence late next year. Financial specifics of the deal remain undisclosed, but both companies are poised to make substantial advancements in the AI sector. OpenAI has been actively expanding its capabilities, clinching significant deals with industry giants such as Nvidia, Oracle, and AMD. CEO Sam Altman emphasized the escalating complexity of AI systems, indicating that a comprehensive approach is necessary to meet the demands of advanced intelligence. The new systems will integrate networking, memory, and computing, all meticulously tailored for OpenAI's specific workloads, leveraging Broadcom's Ethernet technology. By creating its own chips, OpenAI aims to reduce computing costs and maximize its infrastructure investments. Industry estimates suggest that establishing a 1-gigawatt data center could require around $50 billion, with a substantial portion allocated for chip expenses, particularly given Nvidia's pricing. Altman expressed optimism about the partnership's potential to deliver a vast array of computing resources necessary for advanced intelligence applications. He noted that this collaboration could yield significant efficiency improvements, resulting in faster and more cost-effective models. Broadcom has emerged as a key player in the generative AI landscape, supplying custom AI chips, referred to as XPUs, to major hyperscale clients. Although the company does not publicly identify its large-scale customers, analysts have pointed to Google, Meta, and ByteDance as early adopters of its technology. With shares of Broadcom up 40% for the year and a market capitalization exceeding $1.5 trillion, the company is well-positioned to capitalize on the AI boom. OpenAI President Greg Brockman revealed that the company is utilizing its models to enhance chip design efficiency, achieving significant reductions in area requirements. In a conversation about the partnership, Broadcom CEO Hock Tan highlighted the necessity of advanced computing capacity to keep pace with the evolution of frontier models. He asserted that controlling chip production is crucial for long-term success in the AI field. Altman also mentioned that the 10 gigawatts of capacity is merely a starting point, predicting an insatiable demand for high-quality intelligence delivered rapidly and affordably. Currently, OpenAI operates with just over 2 gigawatts of compute capacity, which has been sufficient for scaling its flagship product, ChatGPT, and launching new services like Sora. However, as the demand for AI capabilities grows, OpenAI has secured approximately 33 gigawatts of compute commitments in recent weeks across its partnerships. The landscape of AI is evolving rapidly, and these developments signal a transformative era for both Broadcom and OpenAI as they work towards redefining the boundaries of artificial intelligence.