The Rise of ASICs Is Rewriting the Rules for AI Chip Gaming

On June 8, chip stocks rebounded. Leading gainers were not only traditional GPU giants like NVIDIA, but also leading companies in the ASIC (application-specific integrated circuit) sector such as Marvell Technology and Broadcom. Marvell Technology rose more than 9%, while Broadcom rose more than 3%. Behind this is an ongoing structural shift: ASICs are rapidly challenging GPU dominance.

A report from market research firm TrendForce states that the 2026 ASIC market growth rate is expected to reach 44.6%, far exceeding GPU growth of 16.1%. However, in 2024 and 2025, GPU growth still outperformed ASIC. The appearance of this growth acceleration inflection point means that ASICs have grown from “niche substitutes” into one of the main players in the AI compute market.

Why is ASIC growth so fast? The key lies in cost-effectiveness. Against the backdrop of AI compute shifting toward the inference side and rising demands for cost control in data centers, ASIC chips—thanks to their clear economic advantages—are gradually breaking the GPU monopoly landscape. Google’s in-house TPU is among the most successful examples. Google has opened this compute leasing service to enterprises such as Apple and Meta.

Looking ahead to 2026 to 2027, multiple next-generation ASIC platforms will enter mass production or deployment phases. Platforms such as Google’s new-generation TPU and Amazon’s new-generation Trainium will be shipped in batches. TrendForce predicts that in 2026, ASIC market growth will significantly outpace GPUs. This means the “cake” of the entire AI chip market will continue to expand, rather than simply a contest over existing market share. In the future, the distribution of excess returns in the AI chip sector is likely to be broader—not concentrated in a single GPU leader.

#美股AI概念股普涨