HBM4 Competitive Landscape: Samsung, SK Hynix, and Micron Vie for Market Share
Taylor Wilson
The global HBM market is set to shift from HBM3E to HBM4 in H2 2026. SK hynix leads with a 58% revenue share, Samsung is first to mass-produce, and Micron targets 20% share by year-end — the three-way race will reshape AI memory pricing power.
Who leads right now, and by how much?
Counterpoint Research puts SK hynix at 58% revenue share of the global HBM market in Q1 2026. Samsung and Micron tie at 21% each.
But Q1 revenue is still dominated by the older HBM3E; HBM4 shipments won't show up in volume until H2. This means → current share is the scoreboard for the old race, not the new one.
Kyobo Securities analyst Choi Bo-young forecasts SK hynix will hold over 60% of the HBM4 market, maintaining its position as Nvidia's largest supplier.
How did Samsung grab the first-mover slot?
Samsung began mass-producing HBM4 in February 2026 and crossed $1 billion in related revenue roughly 130 days later. In plain terms = it was the first to move HBM4 from the lab to the factory floor and book real sales.
The edge rests on two in-house technologies: the 1c DRAM process (the most advanced DRAM manufacturing node today) and a 4 nm base die — the logic chip at the bottom of an HBM stack — built in Samsung's own fabs.
Samsung has allocated roughly half its HBM capacity to HBM4. Its next-generation HBM4E chip reportedly exceeds 70% yield, but is still in reliability testing. This means → the first-mover advantage is real, but sustaining the lead depends on ramping the next generation.
Can Micron catch up?
Micron disclosed HBM4 revenue above $1 billion in its latest fiscal quarter. Its 12-layer HBM4 stack is ramping at twice the speed of the 12-layer HBM3E.
Micron targets an HBM market share of roughly 20% by year-end. Its data-center unit posted quarterly revenue above $25 billion, implying an annualized run rate over $100 billion.
Micron expects tight supply-demand conditions to persist beyond 2027. This reflects confidence rooted not just in its own ramp but in a market-wide shortage.
The buyer base is changing — what does that mean?
HBM4 buyers are expanding beyond Nvidia GPUs to custom ASICs — application-specific integrated circuits designed for particular AI workloads — built by Google, Amazon, and Microsoft.
In plain terms = HBM used to go almost exclusively to Nvidia. Now the cloud giants are designing their own chips, and those chips need HBM too. More buyers, same tight supply.
This structural shift is expected to accelerate shipment growth and push prices higher in an already constrained market.
How long can prices keep rising?
Citi Research data show Q2 2026 average selling prices for DRAM and NAND rose 44% and 53% quarter-on-quarter. Nomura expects Q3 prices to climb another 24% and 25%, respectively.
AI data-center investment is the core driver. Nomura forecasts this spending will grow at a 48% CAGR, from $466 billion last year to $33.79 trillion by 2030.
JPMorgan flags sustainability of AI memory spending as the biggest risk in this upcycle: AI memory accounts for roughly 52% of cloud capex this year, expected to exceed 70% next year. This means → the bull case is solid in the short term, but if AI capex slows, the entire chain feels pressure simultaneously.
What should we watch in H2?
Actual HBM4 shipment volumes in H2 will be the key verification point for next-stage market share among the three vendors.
Samsung must prove first-mover production translates into sustained share. SK hynix must defend its primary-supplier status with Nvidia. Micron must deliver on its 20% target.
Put simply = all three have shown their hands. H2 shipment data is when the cards are turned over.
Content is for reference only, not financial advice.