Recent reports suggest that Alphabet (NASDAQ: GOOGL) is collaborating with Advanced Micro Devices (NASDAQ: AMD) on its next-generation tensor processing units (TPUs). According to Wedbush Securities, if this partnership materializes, it could trigger a seismic shift across the semiconductor landscape.

Wedbush analyst Matt Bryson noted in a research report that should Google opt for AMD as a partner over Broadcom, Marvell Technology, or Intel—the latter having recently gained traction in similar transformation projects—it would signal a noteworthy development. Bryson added that the mere rumor underscores the growing importance of ASIC design capabilities and IP blocks, highlighting that in an era of increasingly constrained hardware resources, competition in chip manufacturing is not a zero-sum game.

Semiconductor research firm SemiAnalysis suggests that Google plans to leverage AMD's expertise in CPU design, as well as its advancements in packaging and networking, to develop its tenth-generation TPU. Google unveiled two eighth-generation TPUs in April: the TPU v8t, designed for AI model training, and the TPU v8i, optimized for inference operations.

It is worth noting that while traditional large language model (LLM) training remains dominated by accelerators, the growing demand for inference tasks and reinforcement learning in agentic AI is increasing reliance on general-purpose computing resources. Consequently, some analysts speculate that Google may aim to create a tenth-generation TPU specifically tailored for reinforcement learning and agent workloads, with AMD supplying certain building blocks.

In pre-market trading on Tuesday, shares of Alphabet slipped nearly 1%, while Advanced Micro Devices fell over 3%.