Ray Wang on How AI Is Causing DRAM Prices to Surge - Odd Lots Recap

Podcast: Odd Lots

Published: 2026-02-16

Duration: 45 minutes

Guests: Ray Wang

Summary

The surge in DRAM prices is driven by the increasing demand from AI technologies, which require significant memory resources. This demand is causing supply constraints and price hikes, with no immediate relief in sight.

What Happened

The DRAM industry, traditionally characterized by modest growth and stable prices, is experiencing a dramatic shift due to the booming demand driven by AI. Memory chip prices have risen significantly, impacting major companies such as Apple and Nintendo, which now face increased costs and potential supply restrictions.

In late 2022, spot DRAM prices surged, a trend that has continued as AI applications, particularly those involving High Bandwidth Memory (HBM), require more memory. HBM is more wafer-intensive, producing fewer bits per wafer, yet its high bandwidth is crucial for AI model scaling, thus driving its demand.

The COVID-19 pandemic initially led companies to purchase large quantities of DRAM, creating a temporary upcycle. However, the conservative investment during subsequent down cycles has limited current supply, exacerbating the shortage as AI demand accelerates.

This shortage is further intensified by clean room constraints that restrict the ability to expand DRAM production capacity. While node migration offers a potential solution to increase DRAM output, it is a gradual process and not an immediate fix.

Micron is one of the companies attempting to address this supply issue by expanding its capacity through new fabs and node migration. Despite these efforts, the shortage is expected to persist until at least the second half of 2027, marking a longer than usual cycle of four years compared to the typical 15 to 18 months.

Chinese memory producers are struggling to catch up with Korean giants like Samsung and Hynix, being approximately three to four years behind in technology. Meanwhile, China's market constitutes about 25% of global DRAM demand, underscoring the significance of these developments in the memory chip industry.

The memory boom is likened to a commodity supercycle, with preemptive purchasing behavior from companies anticipating continued AI development. Although DRAM cost increases do not majorly impact capital expenditures for hyperscalers like Microsoft or Amazon, they underscore the critical role of memory in the burgeoning AI landscape.

Key Insights

• The demand for High Bandwidth Memory (HBM) is surging because AI models require its high data throughput, despite its inefficiency in wafer usage. This mismatch between demand and production efficiency is pushing DRAM prices up, impacting tech giants like Apple and Nintendo.
• A temporary spike in DRAM purchases during the COVID-19 pandemic, followed by cautious production cycles, has led to a current shortage. As AI demand accelerates, this scarcity is exacerbated, creating a prolonged supply bottleneck.
• Clean room constraints in DRAM manufacturing facilities limit the expansion of production capacity, contributing to the current shortage. While node migration could increase output, it's a slow process, leaving companies like Micron striving to keep up with demand through new fabs.
• Chinese memory manufacturers lag behind South Korean leaders like Samsung and Hynix by several years technologically. Despite China consuming 25% of global DRAM, its domestic industry struggles with the technological gap, highlighting geopolitical tensions in the tech supply chain.

Key Questions Answered

Why are DRAM prices increasing due to AI?

DRAM prices are increasing because AI applications require large amounts of memory, particularly High Bandwidth Memory (HBM), which is essential for model scaling and processing. This demand surge is outpacing supply, driving prices up.

How are Chinese memory producers competing with Korean companies?

Chinese memory producers are currently about three to four years behind Korean companies like Samsung and Hynix in terms of technology. However, they are trying to close this gap to meet the growing demand in the Chinese market.

What are the potential solutions to the current DRAM supply shortage?

Node migration is a potential solution to the DRAM supply shortage, as it allows more bits to be produced from the same wafer capacity. Companies like Micron are expanding capacity with new fabs to help address supply constraints.