应用材料CEO称AI驱动晶圆设备需求为生涯最强、agentic AI单元晶圆用量达GenAI的2-4倍，但内容为两天前投资者日官方展望的二次转述

Applied Materials CEO Gary Dickerson said customer demand to produce more good dies per wafer and rapidly expand wafer output has reached levels unprecedented in his career, driven by the rise of AI computing. Speaking at a global investor conference, he said the company expects revenue from related businesses to grow by more than 20% over the next several years.

As AI computing demand continues to expand, competition in the semiconductor industry is shifting from breakthroughs in individual chip performance to a system-level contest involving wafer capacity, process yields, memory bandwidth, and advanced packaging.
Dickerson said Applied Materials is focusing on three areas of AI-driven equipment growth: advanced logic foundries, DRAM, and advanced packaging. He said the company's accumulated expertise in materials engineering, process equipment, and integration will help it capture the next phase of industry investment. Applied Materials will continue increasing the revenue value generated by each equipment system and its services, with related revenue targeted to grow by more than 20% over the next several years.
Agentic AI lifts wafer demand
Dickerson said Applied Materials began engaging with innovators in data centers and AI about nine years ago and expected AI to become a major growth driver for the semiconductor industry.
On the demand side, AI data centers have changed the primary source of demand for advanced logic chips. Mobile devices were previously a major driver of advanced logic wafer starts, but data-center demand has risen rapidly in recent years and become the main force behind investment.
Dickerson said AI computing architectures are evolving rapidly. The company expects a significant increase in demand for agentic AI computing in 2026, prompting it to raise its outlook for future wafer-fabrication equipment demand. Compared with computing architectures centered on generative AI (GenAI), agentic AI involves more complex task execution and reasoning processes. Chip configurations could also place greater emphasis on CPUs and other computing components.
Using modular computing units in AI data centers as an example, Applied Materials estimates that an agentic AI computing unit will require roughly two to four times as many wafers as a GenAI pod, with the actual increase depending on system configuration. Based on wafer-start requirements, about 85% of the chip demand associated with an agentic AI computing unit will come from advanced logic, DRAM, and advanced packaging.
The industry significance of this estimate is that AI demand is no longer reflected solely in GPU shipments. CPUs, GPUs, networking chips, custom ASICs, and memory components are all increasingly moving to advanced processes, driving process investment across a broader range of chip products.
Dickerson said Applied Materials remains open to different AI computing architectures. Regardless of how GPUs and CPUs are configured within systems, the company can benefit in related markets as more chips adopt advanced processes.
However, fab space, supply-chain components, energy availability, and financing conditions could affect the pace of actual growth in the equipment market. Even with strong AI demand, industry investment must account for the time required to deliver equipment, build fabs, and bring capacity online.
Yield gains lift equipment value
Dickerson said the semiconductor industry has long focused on technological innovation and performance gains. AI customers in particular continue to pursue higher token-generation output per watt and seek to secure design wins early in each new process generation and chip-architecture transition.
However, the industry is now undergoing another major shift, with customers simultaneously raising their requirements for capacity, yield, and production efficiency. When chip demand exceeds supply, yield improvements become even more economically valuable.
For advanced processes, increasing the number of good dies produced from each wafer directly raises the number of deliverable chips. In advanced packaging, manufacturers must reduce the risk of defects during stacking, interconnection, and integration to prevent high-value dies from being scrapped during back-end processing.
This shift is expanding equipment makers' competitive focus beyond simply supplying process equipment to helping customers accelerate R&D, improve yields, and increase mass-production output.
Applied Materials estimates that equipment and service revenue per system will rise by more than 20% over the next several years as R&D investment increases and yields and wafer output improve.