Memory is becoming one of the highest hidden costs in the AI arms race.
Meta Platforms Inc.
On the company's earnings call, CFO Susan Li said the increase reflects "higher component pricing this year," along with added data center costs tied to future capacity needs.
Meta CEO Mark Zuckerberg was more direct, saying "most" of the higher infrastructure forecast was due to higher component costs - "particularly memory pricing."
The comments highlight a growing pressure point for hyperscalers: AI infrastructure is not just about GPUs.
Advanced servers also need huge amounts of DRAM, NAND, HBM and storage capacity.
The rising demand is flowing through to companies such as Micron Technology Inc.
Amazon & Microsoft Also Noted Memory Costs
Amazon.com Inc.
CEO Andy Jassy said, "The cost of these components, particularly memory, has skyrocketed," adding that there is "just not enough capacity for the amount of demand."
Microsoft Corp.
CFO Amy Hood said Q4 capex would rise to more than $40 billion, including roughly "$5 billion from higher component pricing."
She also said calendar 2026 capex of about $190 billion includes roughly "$25 billion from the impact of higher component pricing."
Hood separately noted that Windows OEMs and channel partners were building inventory due to "increasing memory prices."
Two-Sided Trade
SanDisk, Western Digital, and Seagate are among the S&P 500's top year-to-date gainers, as soaring demand and rising memory prices directly benefit the manufacturers.
The tailwind for memory and storage stocks has become a pain for the hyperscalers.
The AI buildout is lifting demand for memory-heavy infrastructure, but the same surge is raising costs for Meta, Microsoft, Amazon and other hyperscalers.
The memory crunch is a two-sided trade: margin pressure for AI spenders and pricing power for memory suppliers.
