AI consumes memory, cheap phones under $100 may disappear from the market.

Memory cost for budget phones now accounts for 64% of total Bill of Materials (BOM); BOM for sub-$200 models has surged 20%–30% year-to-date. IDC warns sub-$100 Android phones could vanish, forecasting a 15% decline in global smartphone shipments in 2026 and a 22% drop in low-price segment sales.

(Recap: Samsung aggressively raises memory prices by 20%! UBS revises quotes upward: DRAM QoQ +32% in Q3, NAND +30%) (Background: Samsung's operating profit surges 18x, beating Nvidia and Apple, but shares tumble over 6% after earnings release)

Table of Contents

Toggle

• Zero-sum Wafer Battle
• Who Gets Sacrificed First?
• The Disappearing Sub-$100 Phone

Xiaomi founder Lei Jun has publicly warned more than once this year: memory costs are skyrocketing. This is no scare tactic. Over the past year, DRAM used in smartphones has risen nearly 70%, and NAND flash has nearly doubled.

The driving force is not inflation or labor shortages, but AI. Data centers' insatiable demand for HBM (high-bandwidth memory, a high-speed spec dedicated to AI chips) is diverting production capacity across the entire memory supply chain away from phone makers.

Zero-sum Wafer Battle

The root cause is that Samsung, SK Hynix, and Micron are concentrating limited wafer capacity on higher-margin HBM production for Nvidia's AI chips.

The result is a zero-sum game: every extra wafer allocated to AI data centers means one less wafer for LPDDR5X (a common memory spec for mobile devices with read/write speeds superior to previous generations) used in mid-range phones. Currently, HBM has already consumed approximately 23% of global DRAM wafer capacity.

How severe is the price surge? From Q1 2025 to Q1 2026, LPDDR4 rose 250%, LPDDR5 rose 220%; just in Q1 2026, overall DRAM prices increased roughly 90% QoQ from Q4 2025.

If memory and SSD are combined, cumulative increases by end-2026 could approach 130%.

Who Gets Sacrificed First?

The answer lies in the structure of the BOM.

The cheaper the phone, the higher the memory share of the BOM: for ultra-budget devices priced below $99, memory costs now account for 64% of the total BOM; for mid-range phones, it falls to 15%–20%; for high-end flagships, only 10%–15%. In other words, models with less brand premium to absorb costs are hit hardest by memory price hikes.

The numbers are already reflected in selling prices: BOM costs for low-price phones under $200 have risen 20%–30% since early 2026, with some entry-level models seeing price increases as high as 50% within a year. IDC even warns that Android phones priced below $100 may permanently disappear from the market.

Phone makers have few options: raise prices directly, or quietly downgrade specs — cutting base model RAM back to 4GB, or bringing back the long-eliminated microSD card slot.

Meanwhile, Samsung, SK Hynix, and Micron are gradually phasing out DDR4 production lines, concentrating resources on advanced processes for AI and server-grade products, exacerbating the already tight supply for the entry-level segment. It's not that manufacturers are unwilling to supply cheap memory; it's that the profit from allocating wafers to AI far exceeds that from a $100 phone.

The Disappearing Sub-$100 Phone

The consequences are already visible in shipment numbers. Market research firms estimate that global smartphone shipments will decline by approximately 15% in 2026, with low-price segment sales likely dropping 22% — a far steeper decline than the overall market.

At the same time, average selling prices are expected to rise 6.9% in the opposite direction, and over a longer horizon the overall increase could reach 13%–14%. As volume shrinks and prices rise, the entry-level models that once occupied the bottom are disappearing.

Xiaomi, OPPO, and vivo have all simultaneously lowered their 2026 shipment forecasts by 10%–15%, and the Xiaomi 17 Ultra has already confirmed a price increase. These adjustments are forced reactions to the cost structure. Compared with the ongoing ramp-up in HBM procurement by AI data centers, phone makers have had no bargaining power in the allocation of wafer capacity from the start.