JPMorgan Chase: AI inference extends server lifecycles, while rising memory prices suppress PC demand

Deep Tide TechFlow message. According to research from the tide, on July 15 JPMorgan Chase reported a significant upward revision to server shipment expectations: for 2026, the growth rate was raised from 15% to 22%, and for 2027 from 8% to 25%. AI inference is the core driving force. Enterprises that deploy AI models require a large number of inference servers. JPMorgan Chase estimates that by 2028, server CPU shipments will rise from 26 million units to 68 million units, of which Agentic AI-related demand will account for 53 million units. On the PC side, demand is constrained by rising memory prices; brands protect gross margins by raising prices, at the cost of sales. For 2026, PC shipments are expected to decline 8%, and consumer PCs are down 14%. Supply bottlenecks remain the constraint: in CPUs, substrates, memory, PCBs, and power components, no link is loose.

In the U.S. stock market, AI server makers such as Dell Technologies, Hewlett Packard Enterprise, and Super Micro Computer continue to benefit. In the components segment, companies such as Arista Networks, Amphenol, Corning, Lumentum, and Micron Technology benefit from the structural trend of value shifting toward components. JPMorgan Chase’s recommendation: the server components segment is better than contract manufacturing, while overall PCs should be avoided.