Memory has grown to nearly two-thirds of AI chip component costs (epoch.ai)

Article summary

The cost of high-bandwidth memory (HBM) has increased significantly, now accounting for 63% of AI chip component costs, up from 52% in Q1 2024. This shift is driven by tight memory supply and rising prices. As a result, HBM spending has grown from $12 billion in 2024 to $32 billion in 2025. This trend is expected to continue, with HBM likely accounting for an even larger share of AI chip component costs in 2026.

Main themes

• AI chip component costs
• Memory pricing trends
• Supply chain constraints
• Tech industry economics
• Component pricing volatility

What commenters say

• The price of RAM has increased dramatically over the past year, with some users reporting price jumps of 4-5 times for the same products.
• The cost of memory is cyclical and has fluctuated over the years, but the current prices are unusually high due to supply chain constraints.
• Some commenters believe that the high demand for newer memory technologies like DDR5 has led to a shortage and price increase for older technologies like DDR3.
• Others argue that the price increase is not just due to supply and demand, but also due to the market power of large corporations and potential price gouging.
• The rise in memory prices is affecting not just consumers, but also businesses and industries that rely on large amounts of memory, such as data centers and AI applications.
• Some users are exploring alternative solutions, such as using older memory technologies or alternative storage solutions like Optane, to mitigate the effects of the price increase.