Back to list
High-Bandwidth Memory Costs Surge to 63% of AI Chip Components as Total Spending Hits $52 Billion
Industry NewsAI ChipsHBMSemiconductor Industry

High-Bandwidth Memory Costs Surge to 63% of AI Chip Components as Total Spending Hits $52 Billion

A comprehensive analysis by Epoch AI reveals a significant shift in the cost structure of artificial intelligence hardware, with High-Bandwidth Memory (HBM) now accounting for nearly two-thirds of total component spending. Between Q1 2024 and Q4 2025, HBM's share of costs for chips designed by industry leaders including Nvidia, AMD, Google, and Amazon rose from 52% to 63%. While logic die costs remained stable, the relative share of advanced packaging and auxiliary components declined. Total component spending is projected to reach $52 billion in 2025, driven by a massive $20 billion increase in HBM expenditures. This trend is forcing major hyperscalers like Microsoft and Meta to significantly increase their capital expenditure forecasts for 2026 to accommodate rising component prices and persistent supply constraints.

Hacker News

Key Takeaways

  • HBM Dominance: High-bandwidth memory (HBM) share of total AI chip component costs grew from 52% in Q1 2024 to 63% by Q4 2025.
  • Spending Explosion: Total component spending for major AI chip designers is projected to jump from $22 billion in 2024 to $52 billion in 2025.
  • Component Cost Shifts: Logic die costs remained stable at 13-14%, while advanced packaging fell from 19% to 15% and auxiliary components dropped from 15% to 9%.
  • Hyperscaler Adjustments: Microsoft and Meta are increasing their 2026 capital expenditure by billions of dollars specifically to account for rising component prices.
  • Future Outlook: HBM is expected to claim an even larger share of costs in 2026 due to tight supply and rising market prices.

In-Depth Analysis

The Rising Financial Weight of High-Bandwidth Memory

According to data insights from Epoch AI, the economic landscape of AI chip manufacturing is undergoing a rapid transformation. High-bandwidth memory (HBM) has emerged as the most significant cost driver in the production of AI hardware. Between the first quarter of 2024 and the final quarter of 2025, HBM's share of total component spending increased from 52% to 63%. This calculation represents a weighted average across the production volumes of major industry players, including Nvidia, AMD, Google, and Amazon.

In absolute financial terms, the growth is even more pronounced. Spending on HBM across these four major designers surged from approximately $12 billion in 2024 to an estimated $32 billion in 2025. This represents a faster year-over-year increase than any other individual component in the AI chip ecosystem. The total component spend for these chips is expected to reach $52 billion in 2025, meaning HBM alone accounts for a staggering $20 billion of the total growth seen between 2024 and 2025.

Stability in Logic and Efficiency in Packaging

While memory costs have skyrocketed, other critical components have seen their relative cost shares stabilize or even decline. Logic dies, the core processing units of the chips, have maintained a remarkably steady share of the total spend, hovering around 13% to 14% throughout the analyzed period. This suggests that while the complexity of logic may be increasing, its cost relative to the total system is being overshadowed by memory requirements.

Conversely, advanced packaging (such as CoWoS) and auxiliary components have seen their shares of the total cost pie shrink. Advanced packaging fell from 19% to 15%, and auxiliary components saw a more dramatic decrease from 15% to 9%. It is important to note that these decreases are in terms of "share of spend" rather than necessarily absolute cost, as the massive increase in HBM spending naturally compresses the percentage share of other categories.

Hyperscaler Capex and Market Forecasts

The rising cost of components is having a direct and measurable impact on the financial planning of the world's largest technology companies. Hyperscalers are already adjusting their capital expenditure (capex) guidance to prepare for a more expensive hardware environment in 2026. Microsoft’s fiscal year 2026 capex outlook has reached $190 billion, a figure that includes approximately $25 billion specifically attributed to higher component prices. Similarly, Meta has raised its 2026 capex range by $10 billion, citing the rising costs of the hardware necessary to power its AI initiatives. As memory supply remains tight and prices continue to rise, HBM is likely to account for an even larger share of the total AI chip cost in the coming year.

Industry Impact

The shift toward memory-heavy cost structures signifies a bottleneck in the AI hardware supply chain. As AI models become more complex, the demand for high-speed data access is outstripping the advancements in logic processing, making HBM the primary commodity in the AI race. This trend grants significant pricing power to memory manufacturers and places a heavy financial burden on chip designers and cloud service providers. The massive capex increases from Microsoft and Meta suggest that the industry is bracing for a sustained period of high hardware costs, which may influence the pace of AI infrastructure deployment and the eventual pricing of AI services for end-users.

Frequently Asked Questions

Question: Why is the cost of HBM rising faster than other AI chip components?

High-bandwidth memory costs are rising due to tight supply and increasing market prices as demand for AI processing power surges. It has become the fastest-growing component in terms of absolute spend, jumping from $12 billion to $32 billion in just one year.

Question: Which companies were included in the Epoch AI cost analysis?

The analysis focused on AI chips designed by four major industry leaders: Nvidia, AMD, Google, and Amazon, with results weighted by their respective production volumes.

Question: How are logic die and packaging costs trending compared to memory?

Logic die costs have remained stable at approximately 13-14% of total component spend. In contrast, the share of spending for advanced packaging fell from 19% to 15%, and auxiliary components fell from 15% to 9% as HBM took over a larger portion of the total budget.

Related News

How Google Duplex Evolved to Call Businesses on Your Behalf in Gemini
Industry News

How Google Duplex Evolved to Call Businesses on Your Behalf in Gemini

A Google Research Blog entry titled “How Google Duplex evolved to also call businesses for you in Gemini” describes its subject as the evolution of Google Duplex to support calls to businesses on a user’s behalf in Gemini. The entry is dated November 20, 2026, and the supplied article text consists only of the label “Generative AI.” The headline identifies the capability and connects it to Duplex, but the available text does not explain how the system evolved, how users request a call, or where the feature is available. The headline establishes the topic, but the supplied material does not provide a technical account or rollout details.

How AI and Lean Are Helping Automate Formal Verification to Fight Software Slop
Industry News

How AI and Lean Are Helping Automate Formal Verification to Fight Software Slop

In Latent Space, Richard MacManus examines formal verification as an emerging way to fight what the headline calls “slop.” The central claim is that proving software correct with mathematics has traditionally required specialist work, and that AI and Lean are now helping automate that process. The distinction is important: the reported change concerns how the work of proving correctness is done, not what counts as a proof. The supplied excerpt presents automation as assistance, without saying that it replaces specialists or completes proofs independently. It also does not specify which parts of the process AI and Lean help automate, how much work they save, or which software has been verified this way.

AI and Human Claim Solutions to the 167-Year-Old Riemann Hypothesis
Industry News

AI and Human Claim Solutions to the 167-Year-Old Riemann Hypothesis

TechCrunch AI reports that AI claimed a solution to the Riemann hypothesis, a mathematical problem the headline describes as 167 years old. A human also put forward a competing solution. The report, written by Tim Fernholz and published on November 20, 2026, describes mathematicians as astonished by the AI-driven result. The supplied excerpt does not explain either proposed solution, identify the AI system or the human claimant, or say whether either claim has been verified. The distinction between a claimed solution and a confirmed one is central to reading the report: the excerpt establishes that two claims are being reported, but it does not establish that either has resolved the hypothesis.