New here? Follow stories to track developments over time. Create a free account to get updates when stories you care about change.
Overview
Modern AI chips can crunch numbers faster than memory can feed them data. On June 18, 2026, SK hynix said it shipped samples of HBM4E, a stacked memory chip built to narrow that gap, to major customers.
The 12-layer part holds 48 gigabytes and moves data at up to 16 gigabits per second per pin. It also draws more than 20% less power than the prior generation. Memory, not just processors, now sets the ceiling on how large an AI data center can grow.
Why it matters
The memory feeding AI processors is now the limit on data-center size. Whoever ships it fastest sets how quickly AI gets cheaper to run.
Questions about this story
No questions yet — be the first to ask.
Key Indicators
Voices
Curated perspectives — historical figures and your fellow readers.
Play
Exploring all sides of a story is often best achieved with Play.
Higher or Lower
A number from this story, against one from elsewhere in the news — guess which is bigger, then keep the chain going. 5 rounds, 3 strikes; a miss costs a strike and resets your streak.
Keyboard: ↓/L lower · ↑/H higher
0 points — sign up to put that on the leaderboard.
Timeline
Order five events from this story, oldest at top. Each in the right slot scores 1 — neighbours within one slot count too. Your previous result — green ✓ for exact slots, yellow ~ for off by one. Cards now in true chronological order.
Sign up to save your score and track a streak across stories.
Connections
Sixteen names from the news. Find the four hidden groups of four. Four mistakes max.
Sign up to keep a daily streak — a new puzzle lands every day.
Exit debate?
Your progress in this debate will be lost.
- 1 Two AI personas square off on this story.
- 2 You predict who'll win each round — correct picks earn XP.
- 3 One crossfire question is yours to fire. Pick it carefully.
Couldn't generate a topic
Select Your Champions
Choose one persona for each side of the debate
DEBATE TOPIC
Choose personas with different perspectives for a more dynamic debate.
Select debater for this side:
No debate personas available right now.
Select debater for this side:
No debate personas available right now.
Who's Got This Round?
Make your prediction before the referee scores
The referee scores both sides on
Round Results
Set the Crossfire
Pick the question both personas must answer in the final round
Debate Oracle! You called every round!
Sharp Instincts! You know your debaters!
The Coin Flip Strategist! Perfectly balanced!
The Contrarian! Bold predictions!
Inverse Genius! Try betting the opposite next time!
XP Breakdown
Prediction History
People Involved
Organizations Involved
South Korean memory maker that holds the largest share of the high-bandwidth memory used in AI systems.
South Korean rival that shipped the first 12-layer HBM4E samples on May 29, 2026, ahead of SK hynix.
Designer of the AI accelerators that consume most of the world's high-bandwidth memory.
US memory maker and the third qualified supplier of high-bandwidth memory for Nvidia's AI chips.
Timeline
September 2025 June 2026
-
SK hynix ships HBM4E samples
Latest QualificationSK hynix ships 12-layer, 48-gigabyte HBM4E samples running at up to 16 gigabits per second per pin. Its shares hit a record high.
-
Samsung samples HBM4E first
QualificationSamsung ships the industry's first 12-layer HBM4E samples to global customers, claiming a lead over SK hynix.
-
HBM4 mass production starts
ProductionSK hynix begins mass-producing HBM4 for Nvidia's Rubin AI chips, with a reported share near 70% of the order.
-
Paid HBM4 samples reach Nvidia
QualificationSK hynix and Samsung deliver paid HBM4 samples to Nvidia ahead of supply contracts for the Rubin platform.
-
SK hynix finishes HBM4 first
MilestoneSK hynix says it is the first company to complete development of HBM4 and ready it for mass production.
Historical Context
2 moments from history that rhyme with this story — and how they unfolded.
The HBM3E qualification race (2023–2024)
As the AI boom took off, SK hynix qualified its HBM3E memory with Nvidia first and became its main supplier. Samsung, the larger memory maker overall, spent months trying to pass Nvidia's tests and fell behind in the most profitable memory segment.
SK hynix captured most of Nvidia's HBM orders and saw profits surge. Samsung's delays cost it share and management changes followed.
SK hynix entered the HBM4 era as the market leader, with roughly 58% share by early 2026.
It shows why passing qualification matters more than shipping the first sample. Samsung sampled HBM4E first this time, but the HBM3E race was decided in the test lab, not at the sample stage.
The memory supercycle and crash (2017–2019)
A shortage of DRAM and NAND sent memory prices soaring, and SK hynix and Samsung posted record profits. Then demand cooled and new supply arrived. Prices fell hard and earnings collapsed within about a year.
Memory makers booked huge profits in 2017 and 2018, then deep declines in 2019.
It confirmed how cyclical memory is: record highs in chip stocks have repeatedly given way to sharp downturns.
SK hynix shares hit a record on the HBM4E news. The supercycle is a reminder that memory demand and pricing can reverse quickly, even for parts in short supply today.
