AI Chip Component Costs: Memory at 63% | Epoch AI | Epoch AI

Latest Topics AI Progress Scaling Software progress Open models Capabilities Math Industry Leading companies Finances Geopolitics Infrastructure Chips Data centers Energy Impacts Adoption and use Economic impact Future of AI See all topics Our work Featured AI Trends & StatisticsData ExplorersCapabilities & Benchmarking Publications Papers & Reports Data Insights Newsletter Podcast See all publications Data explorers Capabilities Models Frontier Data Centers Chip Owners Companies Polling on AI Use See all data explorers Benchmarks by Epoch AI Epoch Capabilities Index FrontierMath: Open Problems FrontierMath: Tiers 1-4 About About Epoch AI Donate Team Careers Consultations For press Transparency Contact

LatestTopicsAI Progress
ScalingSoftware progressOpen modelsCapabilitiesMath
Industry
Leading companiesFinancesGeopolitics
Infrastructure
ChipsData centersEnergy
Impacts
Adoption and useEconomic impactFuture of AI
See all topicsOur workFeaturedAI Trends & StatisticsData ExplorersCapabilities & BenchmarkingPublications
Papers & ReportsData InsightsNewsletterPodcast
See all publicationsData explorers
CapabilitiesModelsFrontier Data CentersChip OwnersCompaniesPolling on AI Use
See all data explorersBenchmarks by Epoch AI
Epoch Capabilities IndexFrontierMath: Open ProblemsFrontierMath: Tiers 1-4
About
About Epoch AIDonate

TeamCareers

ConsultationsFor press

Transparency
Contact

Data Insights Memory has grown to nearly two-thirds of AI chip component costs Data InsightMay 21, 2026 Memory has grown to nearly two-thirds of AI chip component costs

Cite
By Venkat Somala High-bandwidth memory (HBM) has grown from 52% to 63% of total AI chip component spending between Q1 2024 and Q4 2025. These estimates are an average across all AI chips designed by Nvidia, AMD, Google, and Amazon, weighted by production volume. As a share of spend, logic dies stayed roughly flat near 13%, while advanced packaging fell from 19% to 15% and auxiliary components fell from 15% to 9%. In absolute terms, HBM spend across these four designers grew from roughly $12 billion in 2024 to $32 billion in 2025, a faster year-over-year increase than any other component. Enable JavaScript to see an interactive visualization. HBM will likely account for an even larger share in 2026 as memory supply remains tight and prices rise. Hyperscalers are already anticipating this in capex guidance: Microsoft’s $190 billion FY2026 capex outlook includes about $25 billion from higher component prices, while Meta raised its 2026 capex range by $10 billion, citing higher component prices.
Epoch's work is free to use, distribute, and reproduce provided the source and authors are credited under the Creative Commons BY
license.
Learn more about this graph For each AI chip designed by Nvidia, AMD, Google, and Amazon, we estimate the per-chip cost of four component categories: memory (HBM), logic dies, advanced packaging (CoWoS), and auxiliary components. We then multiply those per-chip costs by estimated quarterly production volumes to get total component spending in each category, and compute each category’s share of total component spending per quarter from Q1 2024 to Q4 2025.
We find that memory’s share rose from 52% to 63% over this period, while packaging fell from 19% to 15% and auxiliary components from 15% to 9%. Logic die share stayed roughly constant near 13–14%. Total component spend on AI chips grew from approximately $22 billion in 2024 to $52 billion in 2025, with HBM spending alone accounting for roughly $20 billion of that increase. Data

Component cost estimates are drawn from our AI Chip Components explorer, which builds chip-level bills of materials from financial disclosures, supplier filings, and analyst reports. Four component categories are tracked:

Memory: HBM stacks (HBM3, HBM3e).
Logic: advanced-node logic dies (3 - 5nm).
Packaging: TSMC CoWoS advanced packaging.
Auxiliary: substrate, power delivery, and other non-logic, non-memory inputs.

See the explorer’s methodology documentation to learn more. Analysis

Each unit of a component has some cost uncertainty, including the price of an HBM stack, a logic die, or a CoWoS package. We model each chip’s per-component cost with a 90% confidence interval and the share is a ratio of this component’s cost over the total, so both numerator and denominator are uncertain. We show two bounds on the share:

Range from this component’s cost: the share when this component’s cost lands at its 5th or 95th percentile and the other three components are held at their medians.
Range when all components vary at extremes: the share when this component sits at one extreme of its CI and every other component lands simultaneously at the opposite extreme of theirs.

Q1 2024:

ComponentShareRange from this component’s costRange when all components vary at extremesMemory52%48 - 56%42 - 62%Logic14%12 - 17%10 - 20%Packaging19%14 - 24%12 - 27%Auxiliary15%13 - 18%11 - 21%
Q4 2025:

ComponentShareRange from this component’s costRange when all components vary at extremesMemory63%60 - 67%54 - 73%Logic13%10 - 16%9 - 19%Packaging15%11 - 19%9 - 22%Auxiliary10%8 - 10%7 - 12% Assumptions and limitations

Component costs can vary by contract, supplier, and timing, so our per-chip estimates carry some uncertainty. Our estimates of the number of chips produced each quarter and the chip-type mix also carry uncertainty, both of which propagate into the shares we report. Download this data AI chip component cost shares by quarterCSV, Updated May 21, 2026

Explore this data AI Chip ComponentsAI chip supply chain consumption data. Related topics ChipsFinancesLeading companies Cite

Epoch AI’s work is free to use, distribute, and reproduce provided the source and authors are credited under the Creative Commons Attribution license.CitationVenkat Somala (2026), "Memory has grown to nearly two-thirds of AI chip component costs". Published online at epoch.ai. Retrieved from 'https://epoch.ai/data-insights/ai-chip-component-cost-shares' [online resource]. Accessed 24 May 2026.BibTeX Citation@misc{epoch2026aichipcomponentcostshares,
title={{Memory has grown to nearly two-thirds of AI chip component costs}},
author={Venkat Somala},
year={2026},
url={https://epoch.ai/data-insights/ai-chip-component-cost-shares},
note={Accessed: 2026-05-24}} Investigating the trajectory of AI for the benefit of society Get the latest from Epoch AI in your inbox Subscribe Our Work Latest AI Trends & Statistics Papers & Reports Data Insights Newsletter: Gradient Updates Podcast Data on AI Models Frontier Data Centers Hardware Companies Chip Sales Polling Benchmarking AI Capabilities FrontierMath Company About Team Consultations Transparency Donate Careers © 2026 Epoch AI Privacy Policy Cookie Policy Designed by And—Now

FeedbackFeedback

Have a question? Noticed something wrong? Let us know.MessageIf you would like a reply, please include your name and email address.NameEmail addressCancelSubmit Memory has grown to nearly two-thirds of AI chip component costs High-bandwidth memory (HBM) accounts for 63% of AI chip component costs, up from 52% in Q1 2024. Epoch AI's breakdown of component cost shifts across major chip designers.

High-bandwidth memory (HBM) has become a dominant factor in the cost structure of artificial intelligence (AI) chip components, accounting for nearly two-thirds of total spending. Analysis by Epoch AI indicates that the share of HBM in AI chip component costs increased significantly, growing from 52% in the first quarter of 2024 to 63% by the fourth quarter of 2025. This shift reflects a fundamental change in the economics of AI hardware development across major designers including Nvidia, AMD, Google, and Amazon, weighted by production volume. In contrast, the share of logic dies remained relatively stable, hovering near 13% across the period. Furthermore, spending on advanced packaging, such as TSMC CoWoS, decreased from 19% to 15%, while auxiliary components saw a reduction from 15% to 9%. In absolute terms, HBM expenditure escalated sharply, rising from approximately $12 billion in 2024 to $32 billion in 2025, representing a faster year-over-year increase compared to other component categories.

Epoch AI estimates these component costs by modeling the per-chip cost of four main categories: memory (HBM stacks), logic (advanced-node dies like 3 to 5nm), packaging (CoWoS), and auxiliary components (substrate, power delivery, and other inputs). The methodology involves multiplying estimated per-chip costs by projected quarterly production volumes to determine total spending and calculate the share of each category across the total component expenditure. This analysis inherently models uncertainty, as component costs themselves carry variability, and the calculation of cost shares accounts for ranges based on the statistical distribution of these uncertainties.

For instance, the cost distribution ranges demonstrated considerable volatility. In Q1 2024, the share range for HBM was 48% to 56%, whereas the range for logic dies was 12% to 17%. By Q4 2025, the range for HBM expanded further to 60% to 67%, while the range for packaging narrowed to 11% to 19%. This variance highlights the sensitivity of component cost distribution to market conditions and supply chain dynamics. The uncertainty is further propagated by estimates of chip production volumes and the mix of chip types, all of which influence the reported component shares.

Looking forward, the trajectory suggests that HBM will likely secure an even larger share of AI chip component costs in 2026, driven by persistent memory supply tightness and rising prices. This trend is reflected in the financial planning of hyperscalers; for example, Microsoft's fiscal year 2026 capital expenditure outlook includes approximately $25 billion allocated to higher component prices, and Meta increased its 2026 capital expenditure range by $10 billion due to these escalating costs. This evolving component cost landscape underscores the critical importance of memory infrastructure in the ongoing scaling and deployment of advanced AI systems.