Reading time: 5 min
Samsung Electronics reclaimed the global DRAM market lead in Q4 with $19.3 billion in quarterly sales — its first time on top in three quarters — but the victory arrives at a moment when the memory market itself is breaking down, with AI demand driving the worst shortage in decades and prices doubling inside a single quarter.
The Crown Returns
TrendForce data published this week shows Samsung captured a 36% share of the global DRAM market in the October–December period, with revenue jumping 43% quarter-on-quarter. The company overtook SK Hynix, which posted a 25.2% revenue increase to $17.2 billion but saw its share slip to 32.1%. Micron Technology held third with $12 billion in sales and a 22.4% share. Samsung had lost the DRAM leadership in Q1 2025 — its first surrender of the top spot in 33 years — after falling behind in qualifying its high bandwidth memory chips for Nvidia’s AI accelerators. By Q3, Samsung had narrowed the gap to 0.6 percentage points. By Q4, the combination of successful HBM3E shipments to Nvidia and a surge in legacy DRAM prices put Samsung definitively back on top.
The catalyst wasn’t just technology. Samsung operates the largest DRAM production capacity among the Big Three — roughly 650,000 wafers per month, versus 450,000 for SK Hynix and 300,000 for Micron. When legacy DRAM prices surged 45–50% quarter-on-quarter in Q4, Samsung benefited disproportionately because legacy products still make up a larger share of its revenue mix. Memory gross margins at Samsung and SK Hynix reportedly surpassed TSMC’s for the first time since 2018.
The Shortage Behind the Profits
The headline numbers look like a celebration. The underlying market dynamics look like a crisis. Every major DRAM maker has shifted production toward high bandwidth memory and high-capacity DDR5 for AI servers, and the reallocation has created a structural deficit in conventional memory. IDC warned in February that this is “not just cyclical” but represents “a potentially permanent, strategic reallocation of the world’s silicon wafer capacity.”
Data centres will consume roughly 70% of all memory chips produced in 2026, according to TrendForce. The Big Three have allocated 18–28% of total DRAM production to HBM — and each bit of HBM produced costs three bits of conventional DRAM, because the stacking process is that capacity-intensive. Demand is growing at approximately 35% year-on-year while supply expands at around 23%. The gap is widening.
Prices That Haven’t Been Seen Before
TrendForce’s latest revision, published in early February, raised its Q1 2026 conventional DRAM price forecast from a 55–60% quarter-on-quarter increase to 90–95%. PC DRAM specifically is expected to rise over 100% in a single quarter — a record. Memory that typically accounted for 15–18% of a PC’s bill of materials could now represent 35–40%, according to Compal, one of the world’s largest contract manufacturers. IDC expects average PC prices to jump up to 8%, and the overall PC market could shrink by 9% if conditions don’t improve.
The squeeze extends well beyond PCs. Counterpoint Research warns that DDR5 server memory costs could surge 100%. IDC expects smartphone OEMs to freeze RAM at 12GB rather than upgrading to 16GB. Samsung itself has cautioned about “weak smartphone demand due to memory supply and price impacts” in Q1. Samsung has also hiked its own memory chip prices by up to 60%, according to Reuters. Dell is planning double-digit PC price increases. Valve has confirmed the Steam Deck is out of stock because of memory shortages. Japanese retailers have halted desktop PC orders. Framework has changed its return policy to prevent customers from buying laptops, stripping the RAM and returning the chassis.
The Exit That Made It Worse
Micron’s decision to kill its consumer Crucial brand — with shipments ending February 2026 — removed one of the last direct consumer-facing channels from a major DRAM producer. The company refocused entirely on HBM and enterprise customers. SK Hynix has said its entire 2026 production capacity is already sold. Micron’s chief business officer has admitted the company can meet at most two-thirds of medium-term demand for some customers. New fabs under construction in Idaho won’t produce until 2027–2028. TrendForce’s Avril Wu has said new capacity “won’t make a noticeable difference in global supply until 2028.” The three producers control approximately 90% of global DRAM supply. When all three simultaneously pivot toward the same high-margin product, the rest of the market doesn’t get rationed — it gets starved.
What Breaks First
For investors, Samsung’s DRAM crown is a sideshow. The real story is whether the memory supercycle breaks something before new supply arrives. SK Hynix has predicted the shortage will last through late 2027. Nvidia CEO Jensen Huang has acknowledged that high memory costs may slow GPU adoption beyond 2026. Cisco’s stock had its worst day since 2022 after flagging memory-price pressure. Morgan Stanley slashed ratings on Dell, HP and HPE.
The irony is precise: Samsung, SK Hynix and Micron have never been more profitable, and the devices their products go into have never been harder to buy. Every wafer converted to AI memory is a wafer taken away from a laptop, a phone, a car or a router. The market will self-correct — it always does — but TrendForce, IDC and TechInsights all agree: not before 2028. Until then, the DRAM crown comes with a price tag that the rest of the world is paying.
Sources: Korea Times, TrendForce, CNBC, IDC, EE Times
