SK Hynix Ships HBM4 to Nvidia Ahead of Schedule
Miles Bennett
SK Hynix began shipping 12-layer HBM4 finished products to Nvidia in late June, months ahead of the September mass-production target — yet its stock posted a record 15.4% single-day drop, as the market prices in a gap between delivery timing and actual volume.
Why did Nvidia ask SK Hynix to rush?
The shipment targets Nvidia's next-gen Vera Rubin AI accelerator — final-spec product, not samples.
Nvidia asked in April to compress lead time from the usual ~four months down to ~two months.
This means → Nvidia's urgency to launch next-gen AI chips is so acute it is pulling the supply chain forward.
SK Hynix prioritized Nvidia's orders and streamlined parts of its standard qualification process.
It shipped early — but is there enough volume?
Current shipment volumes remain limited; supply reaches meaningful scale only from September as wafer starts ramp up.
The industry expects HBM4's share in SK Hynix's HBM sales mix to surpass HBM3E in Q4.
In plain terms = the June batch is more like turning in homework early — the real volume test comes in Q4.
How can SK Hynix undercut Samsung on price?
The HBM4 uses 1b-node DRAM dies — 1b is among the most advanced DRAM manufacturing nodes today.
The base logic die is built on TSMC's 12 nm process, a combination that lowers overall cost.
SK Hynix has already completed 2027 HBM4 pricing negotiations with Nvidia, locking in its advantage.
This means → process-node choice directly drives pricing power; SK Hynix trades a mature foundry node for cost headroom.
What does the market-share picture look like?
Counterpoint Research data: SK Hynix held 58% of HBM revenue in Q1, with Samsung and Micron at 21% each.
Industry sources expect SK Hynix to remain Nvidia's largest HBM4 supplier at ~60%, with some estimates as high as 70%.
SK Hynix also began shipping 8-layer HBM4 to AMD in July, broadening its customer base.
Why did the stock crash on good news?
On July 13 SK Hynix fell 15.4% in Seoul — its largest single-day drop on record.
Reuters attributed the sell-off to profit-taking after the company's Nasdaq listing and cooling earnings expectations.
NH Investment Securities analyst Ryu Young-ho noted investors had expected HBM4 shipment volumes to rise significantly from Q2, but actual volumes fell short.
This reflects a market that prices volume, not timing — faster delivery without enough scale still triggers a sell-off.
Can Samsung close the gap?
Samsung announced HBM4 mass production and commercial shipment as early as February, ahead of SK Hynix.
On May 29 Samsung began shipping 12-layer HBM4E samples using 1c-node DRAM and a 4 nm foundry logic die, offering over 20% speed improvement over its HBM4.
In plain terms = Samsung is betting a higher-spec HBM4E to seize the premium tier, but SK Hynix holds Nvidia's bulk orders and a cost edge — whether Q4 volume ramp delivers as promised will decide this round.
Content is for reference only, not financial advice.