Marvell pops 6% on report it will help Google design two new custom AI chips. Broadcom shares sink

Marvell Technology saw its stock jump nearly 6% on Monday following reports that Google plans to engage the chip design firm for the development of two AI-focused chips. Previously, Google had been depending on its competitor Broadcom for the creation of its proprietary Tensor Processing Units (TPUs). In contrast, Broadcom's shares dipped by almost 2% in response to the news from The Information. The anticipated partnership between Google and Marvell is expected to encompass a TPU and a memory processing unit, according to the report. Both companies play a crucial role in transforming chip designs into silicon, providing essential backend support before these processors are manufactured at large fabrication facilities like those operated by Taiwan Semiconductor Manufacturing Company. This collaboration has been instrumental in the growth of both Marvell and Broadcom, particularly as tech giants increasingly seek to create their own AI accelerators. It comes as no surprise that Google is looking to broaden its chip partnerships beyond Broadcom, especially given the escalating demand for AI technology. Notably, Google and Broadcom recently renewed their collaboration, extending their partnership through 2031. Moreover, Meta announced a significant deal with Broadcom last week, committing to use 1 gigawatt of its custom MTIA chips that leverage Broadcom technology. Marvell's stock has experienced a remarkable surge, climbing over 20% in March after the company reported strong fourth-quarter earnings amidst rising AI demand. The stock has continued its upward trajectory, increasing nearly 50% in April alone. Additionally, Nvidia's recent $2 billion investment in Marvell facilitates greater access for Nvidia customers to the application-specific integrated circuits (ASICs) produced by major players like Google. Since launching its first TPU in 2015, Google has been a pioneer in developing custom ASICs for AI workloads, with other tech giants like Amazon, Meta, Microsoft, and OpenAI following suit. Google unveiled its latest 7th generation 'Ironwood' TPU in November and is expected to introduce additional chips during its upcoming AI conference, Google Cloud Next. Although originally intended for internal tasks, Google’s custom microchips have been accessible to cloud clients since 2018, with companies such as Meta, Anthropic, and Apple now utilizing TPUs. The AI chip sector has been grappling with memory shortages from key suppliers like Micron, SK Hynix, and Samsung, creating challenges for manufacturers in meeting the growing demand for AI technology.