Alphabet Diversifies AI Chip Supply Chain with Potential Marvell Partnership

Alphabet Inc. is seeking to expand its custom silicon capabilities by negotiating a partnership with Marvell Technology. According to reports, the tech giant aims to develop two new custom AI chips to support its growing artificial intelligence infrastructure and cloud workloads. Historically, Google has collaborated closely with Broadcom to design its Tensor Processing Units (TPUs). However, the shift toward Marvell suggests a strategic move to diversify its supply chain and mitigate dependency on a single vendor as the demand for AI compute scales globally. The proposed collaboration involves the design of a new TPU for running AI models and a complementary memory processing unit. The latter is intended to enhance the performance of TPUs by bringing memory closer to AI accelerators, potentially utilizing Marvell's Structera memory controller technology to boost bandwidth and capacity. The companies reportedly aim to complete the memory unit's design by next year before moving to test production. Marvell has already seen significant momentum, with shares rising 95% in 2026 and 227% over the last year. Company management has indicated an 'unprecedented level of activity' with hyperscalers, noting that over 20 chip designs are slated for production by fiscal 2029. This expansion into Alphabet's ecosystem positions Marvell as a critical player in the hyperscale AI hardware race.