Samsung Electronics is pushing its major customers to accept up to a 20% increase in DRAM memory prices for the third quarter of 2026, marking the company’s third consecutive quarterly hike as demand for AI server memory continues to outpace global manufacturing capacity. If buyers comply even partially, the effects will ripple through smartphones, laptops, and cloud data centers worldwide.
What Happened
According to industry research firm TrendForce and multiple reports published in early July, Samsung entered negotiations with its major customers — including smartphone makers, cloud server operators, and PC manufacturers — seeking a 20% quarter-over-quarter increase in DRAM average selling prices. LPDDR memory, used in mobile devices and thin laptops, faces particularly acute supply shortages and could see increases exceeding 20%. Samsung has already shared its pricing targets with clients, and memory-intensive industries are bracing for another wave of cost increases.
TrendForce analysts expect actual contracted price increases to land between 13% and 18% quarter-over-quarter — somewhat below Samsung’s stated target — due to softer demand from consumer electronics segments providing customers with some negotiating room. Still, even a 13% increase represents a significant jump on top of two already-elevated quarters.
Why It Matters
Price hikes of this magnitude flow downstream to consumers within one to two quarters. PC and laptop manufacturers, smartphone OEMs, and cloud data center operators all depend heavily on DRAM, and their cost increases tend to appear in retail pricing or squeezed margins. For consumers, a memory-intensive device that cost a certain amount in 2025 may cost meaningfully more by late 2026 if Samsung’s targets take hold.
The cumulative impact of three consecutive quarterly hikes is already historic. Samsung raised prices roughly 90% in Q1 2026 and a further 50–60% in Q2. Combined, the effect has more than doubled DRAM selling prices from their late-2025 levels. The sustained pricing power reflects a fundamental supply-demand imbalance that artificial intelligence infrastructure has significantly amplified, as hyperscale cloud providers race to expand AI server capacity and redirect manufacturing resources away from consumer-grade memory. Google’s AI buildout drove a 37% surge in electricity demand last year, underscoring how aggressively the tech giants are expanding their AI infrastructure — and their memory requirements along with it.
Background and Context
The DRAM pricing cycle has been running hot since late 2025, driven by booming orders from hyperscale cloud providers prioritizing AI deployment. Samsung is not alone in this market: Micron and SK Hynix are the two other major producers, and both face similar capacity constraints as high-bandwidth memory (HBM) for AI accelerators competes with commodity DRAM for the same fab resources.
Samsung’s role in the broader AI chip ecosystem is also expanding beyond memory. Anthropic is currently in preliminary talks with Samsung to manufacture a custom 2nm AI inference chip, a project that would deepen Samsung’s involvement in the global AI supply chain well beyond its memory business. Separately, Chinese AI lab DeepSeek is designing its own inference chip to reduce dependence on Nvidia, adding further demand pressure to the broader semiconductor ecosystem as more AI labs seek custom silicon.
Samsung’s own financial results illustrate how profitable the AI memory boom has become. The company’s preliminary Q2 2026 earnings showed operating profit of approximately 89.4 trillion won — roughly 19 times higher than the same period a year earlier. The company is also reportedly evaluating a dual listing on a US stock exchange, which would give American investors direct access to its AI-driven earnings.
What Comes Next
Buyers have limited alternatives. Micron and SK Hynix face equivalent capacity constraints, meaning customers cannot easily switch suppliers to avoid Samsung’s price demands. Some buyers may delay large purchases, lock in longer-term supply agreements at pre-hike prices, or absorb the increases and pass them along to their own customers.
The broader question for the industry is whether AI-driven memory demand can sustain elevated pricing through the end of 2026 and into 2027. If AI server buildout slows — due to policy changes, capital constraints, or a plateau in model scaling — memory oversupply could return relatively quickly and pricing could reverse sharply. For now, however, the trajectory points firmly upward, and Samsung’s pricing ambitions reflect a rare moment of sustained leverage for a memory market long characterized by brutal boom-bust cycles.
