← Retour au blog
tech 24 May 2026

Memory: The New Heavyweight in AI Chip Costs

Memory now accounts for nearly two-thirds of AI chip component costs. An analysis of trends and implications for tech giants.

Introduction

In the world of artificial intelligence, chips are the engines driving innovation. Today, a particular component of these chips is taking on unprecedented importance: memory. According to a recent analysis by Epoch AI, memory now accounts for 63% of AI chip component costs, a significant jump from 52% in 2024.

Why is Memory Taking Center Stage?

Memory, especially high-bandwidth memory (HBM), is crucial for AI chip performance. It enables the rapid and efficient processing of the vast amounts of data required to train complex AI models. This growing need translates into rising demand and prices, pushing the share of memory costs to new heights.

Key Figures

Between 2024 and 2025, HBM spending for the four major chip designers—Nvidia, AMD, Google, and Amazon—increased from $12 billion to $32 billion. This rise reflects not only increased unit costs but also the growth in production volume to meet escalating demand.

Implications for Tech Giants

The surge in memory costs has significant implications for tech giants, particularly in terms of financial planning. Hyperscalers like Microsoft and Meta have already adjusted their capital expenditure forecasts for 2026, anticipating significant component price increases.

Example: Microsoft and Meta

Microsoft announced a $190 billion fiscal year 2026 budget, with $25 billion attributed to higher component prices. Meanwhile, Meta increased its investment budget by $10 billion, citing increased component costs.

The Future of AI Chip Costs

With forecasts indicating continued cost increases, chip manufacturers will need to innovate to optimize memory efficiency. Solutions may include improving packaging technologies or developing new chip architectures.

Innovation and Strategies

Companies may need to turn to emerging technologies such as non-volatile memory or integrated memory to reduce costs. Additionally, optimizing software to better utilize existing memory could offer efficiency gains.

Conclusion

The skyrocketing costs of memory in AI chips present both a challenge and an opportunity. To remain competitive, industry players must not only anticipate these costs but also explore new technological avenues to manage them.

Let's discuss your project in 15 minutes.

AI chips memory cost HBM tech giants innovation
Deepthix newsletter · 100% AI · every Monday 8am

An AI agent reads tech for you.

Our AI agent scans ~200 sources per week and ships the best articles to your inbox Monday 8am. Free. One click to unsubscribe.

Visit the newsletter page →

Want to automate your operations?

Let's talk about your project in 15 minutes.

Book a call