Soaring AI computing demand has sent memory chip prices climbing sharply, with upstream price pressures now cascading down to end-user devices. In recent moves, Amazon.com has raised prices across multiple hardware products, following similar actions by Apple, Microsoft, and Dell, which have all increased product pricing or trimmed standard memory configurations to offset rising chip costs.

Market research firm IDC has previously warned that the global memory shortage could persist through 2027, with industry forecasts suggesting memory chip prices may peak in 2028 before gradually stabilizing. Amazon recently announced price hikes on several hardware devices, including its Echo smart speakers, Kindle e-readers, and Eero router lineup, with some products seeing increases of up to 60%. The 16GB Kindle jumped from $109.99 to $149.99, while the entry-level Echo Dot rose from $49.99 to $79.99—a 60% surge and the steepest increase in this round.

Amazon attributed the adjustments to sharply rising memory and storage component costs across the consumer electronics sector, noting that the company had absorbed as much of the cost pressure internally as possible before resorting to across-the-board price changes. The company also indicated it would continue offering periodic promotional discounts over the next year to provide consumers with some relief.

Where the price pressure starts

Amazon is hardly the first company in the consumer electronics space to raise prices due to storage cost inflation. Apple and Microsoft had already increased prices on select hardware products earlier in response to memory component cost pressures. On August 24, renowned tech journalist Mark Gurman reported that Apple plans to raise prices on its iPhone 18 series launching in September, with increases essentially locked in due to memory and chip supply constraints driving up overall bill-of-materials costs.

On August 25, TrendForce released a forecast projecting that global tech giants will pour $1.383 trillion into AI infrastructure capital expenditures next year, with nearly 70% of that AI infrastructure investment cost flowing into DRAM and NAND flash memory. Analysts note that AI servers' surging memory capacity demands, combined with sharp chip price increases, have shifted the global AI industry's competitive battleground from GPU supremacy to a scramble for memory supply.

TrendForce data shows that contract prices for server DRAM used in AI servers have already risen 64% cumulatively in the second half of 2025, with an additional approximately 270% increase expected this year. Enterprise-grade SSDs used for massive data storage saw prices jump 35% in the second half of last year, with cumulative gains of roughly 235% projected for this year. TrendForce predicts that memory chip contract prices will remain elevated through 2027.

High-bandwidth memory (HBM) prices are also on an upward trajectory. Analysts point out that while general-purpose memory price increases may be limited following long-term supply agreements signed in Q2, HBM prices could surge between 70% and 140% next year amid severe supply shortages.

Why just a handful of key plays matter

Guotai Haitong's research indicates that the storage industry has transformed from a beneficiary of AI computing investment into a critical bottleneck constraining AI infrastructure buildout. The structural shortage driven by AI demand runs through the entire supply chain, with the supply-demand gap still widening. On pricing, the slowdown in Q3 2026 contract price growth is not a sign of cycle peaking—the uptrend is expected to persist through 2027. Long-term agreements are not limiting price upside but rather making price trends more predictable, trading short-term price flexibility for long-term earnings and cash flow visibility. Market attention is shifting from single-quarter price/profit elasticity toward multi-year high profitability sustainability and capital returns supported by strong cash flow generation, which should drive valuation multiples higher.

On the supply side, the brokerage believes supply discipline remains underappreciated by the market. Most new capital expenditure from memory manufacturers is directed toward new fab construction and cleanroom buildouts, with capacity expansion plans progressing in a phased manner. Actual new capacity additions in 2026-2027 will be limited. Structurally, the industry has already fully priced in weak consumer terminal demand, while data center AI demand remains robust. Under flexible capacity allocation, the marginal impact of traditional consumer terminals on industry supply-demand dynamics is diminishing, and this will not alter the supply-demand imbalance or the trend of moderate price increases.

Tianfeng Securities points out that this round of memory industry improvement is not driven by single-factor inventory replenishment at terminals, but rather by the combined effect of demand structure changes, rational supply expansion, and supply chain inventory cycle repair. As AI computing infrastructure buildout continues, the importance of high-end storage demand is further elevated. According to Omdia, server DRAM demand share is projected to rise from 50% in 2025 to 71% by 2030. Frost & Sullivan data shows the global AI endpoint device storage market is expected to grow from $39.5 billion in 2025 to $300.8 billion by 2030.

On the supply front, the three major memory manufacturers have recently maintained capital expenditure ratios mostly in the 23% to 35% range, with no aggressive capacity expansion observed. Some capacity is being preferentially allocated to high-margin products like HBM and large-capacity enterprise SSDs. Against the backdrop of demand structure upgrades and persistent supply constraints, memory price levels are expected to find fundamental support. Investment focus should remain on memory modules, memory controller chips, packaging and testing, as well as upstream equipment and materials.

Related stock picks

SMIC (00981): SMIC's core product portfolio spans multiple areas, including logic chips, memory chips, and analog chips. In late August, DBS released a report noting that SMIC's Q2 revenue reached $3.01 billion, up 36% year-over-year and 20% quarter-over-quarter, exceeding market expectations by 5%. Gross margin came in at 25.3%, surpassing market consensus by approximately 4 percentage points, benefiting from higher average selling prices, improved capacity utilization, and a better product mix. DBS raised its earnings forecasts for SMIC for 2026-2028 by 13.4% to 16.5%, lifted its target price from HK$90 to HK$96, and maintained a "Buy" rating. Management guided Q3 revenue growth of 2% to 4% quarter-over-quarter, representing roughly 30% year-over-year growth at the midpoint, broadly in line with market expectations. Gross margin guidance of 26% to 28% came in approximately 4.3 percentage points above market consensus.

HUA HONG GRACE (01347): Hua Hong Semiconductor reported 2026 interim results with sales revenue of approximately $1.378 billion, up 24.5% year-over-year, and wafer shipments up 17.9%, hitting an all-time high. Gross profit reached $204.5 million, up 83.2% year-over-year, while profit attributable to parent company shareholders stood at $59.568 million, a 409.0% surge. The company attributed the record sales to increased wafer shipments and higher average selling prices. Gross profit growth was driven by improved ASPs and cost efficiency measures, partially offset by higher depreciation expenses. On capacity expansion, all process equipment required for the 83K capacity at the Wuxi Phase II project (Fab 9) had been moved in as of end-June 2026, with installation and commissioning underway and planned capacity targets expected by the end of Q3. Concurrently, the company is advancing its acquisition of a 97.5% stake in Huali Micro, which passed the review in mid-June, with completion expected in Q3. Upon consolidation, the company will further enhance its 12-inch wafer foundry capacity, with complementary technology platforms enabling broader application coverage and a more comprehensive technology portfolio to serve customers with diverse solutions.

GIGADEVICE (03986): The company's memory chip products are seeing simultaneous volume and price increases, driving a substantial improvement in profitability. In the first half of 2026, memory chip revenue reached RMB 9.827 billion, up 245.44% year-over-year, with gross margin at 67.57%. Niche DRAM and SLC NAND Flash products posted rapid year-over-year growth. MCU products benefited from demand pull from industrial, consumer, and automotive sectors, with shipment volumes achieving solid growth. For the period, the company reported total revenue of RMB 11.566 billion, up 178.67% year-over-year, and net profit attributable to parent of RMB 6.857 billion, up 1,091.50%.

Montage Technology (06809): As a leader in the memory interconnect chip market, Montage Technology is transitioning from a memory interconnect leader to a platform-type interconnect chip company. In the AI era, leveraging its long-standing focus on high-speed interconnect chips between computing and storage, the company has built a rich product matrix and holds a leading global market position. Its PCIe Retimer products have entered the volume ramp-up phase, and the company is extending along the same technology path into the higher-value PCIe Switch segment, potentially opening new growth avenues. As AI shifts from model training to inference, particularly with the development of AI agents, demand for server CPUs continues to rise, making the company a primary beneficiary of robust CPU demand.