Memory costs have surged to nearly two-thirds of total AI chip component expenses, reaching 63% by the end of 2025, according to epoch.ai. This marks a significant increase from 52% in early 2024, driven primarily by high-bandwidth memory (HBM) spending, which rose from $12 billion in 2024 to $32 billion in 2025.
The data, averaged across AI chips designed by Nvidia, AMD, Google, and Amazon and weighted by production volume, shows that while memory costs climbed sharply, spending on logic dies remained stable at around 13%. Meanwhile, advanced packaging costs decreased from 19% to 15%, and auxiliary components dropped from 15% to 9%. The rapid growth in HBM expenditure outpaced all other components, reflecting tight memory supply and rising prices in the market.
This trend matters because memory has become the dominant cost driver in AI chip production, influencing capital expenditure decisions by major hyperscalers. For example, Microsoft’s fiscal 2026 capital expenditure outlook of $190 billion includes about $25 billion attributed to higher component prices, while Meta increased its 2026 capex range by $10 billion. These shifts highlight the critical role of memory supply constraints and price inflation in shaping the AI hardware landscape.
Looking ahead, memory is expected to account for an even larger share of AI chip costs in 2026 as supply remains tight and prices continue to rise. Industry watchers should monitor how these cost dynamics affect chip design strategies and capital spending plans among leading technology companies, potentially influencing the pace and scale of AI hardware deployment. (epoch.ai)