HBM4
HBM4 high-bandwidth memory is in commercial production, but each customer has to qualify it separately. Micron reports high-volume shipments for its lead customer's platform and samples for other customers. Samsung separates its HBM4 production from HBM4E sampling. We do not know that every supplier is qualified for every Rubin configuration, or that price premiums will fall on schedule.
Research snapshot · updated Nov 12, 2026
Core thesis
HBM4 roughly doubles per-stack bandwidth over HBM3E and is the memory feeding the next AI accelerator generation. It is no longer a development bet: SK Hynix hit its production milestone first in February 2026, and at GTC Taipei in June 2026 Nvidia confirmed all three makers – SK Hynix, Samsung and Micron – qualified for the Vera Rubin platform, which entered full production the same month. The live question is economic, not technical: who ramps 12-high yields fastest and how quickly SK Hynix’s launch-quarter allocation lead and pricing premium erode now that three suppliers are shipping.
State of the art (2026)
HBM4 is in production. At GTC Taipei on 5 June 2026 Nvidia confirmed all three memory makers – SK Hynix, Samsung and Micron – are qualified to supply HBM4 for the Vera Rubin platform, which entered full production the same month. SK Hynix, first to its production milestone in February 2026, holds roughly 60–70% of initial Rubin allocation; Samsung takes 25–30% and Micron the remainder. The differentiator is speed: the JEDEC base runs 8 Gbps per pin, but Nvidia specified 11.7 Gbps to feed Rubin, whose first GPU pairs 288GB of HBM4 across eight stacks. The supply question is no longer who can make HBM4, but how fast each can ramp 12-high yields without eroding the launch-quarter premium.
Explore the deeper analysis in CanaryIQ
Explore what is covered below. These previews show the structure; log in to read the full analysis.
Signal stack
Evidence stacked leading → lagging
9 signals
Signal stack
Evidence stacked leading → lagging
Technology-native KPIs
Metrics that predict trajectory, tracked over time
3 tracked
Technology-native KPIs
Metrics that predict trajectory, tracked over time
Landscape map
Who builds what, and who depends on whom
53 players · 6 layers
Landscape map
Who builds what, and who depends on whom
Catalyst calendar
Dated events that will move the position
4 ahead
Catalyst calendar
Dated events that will move the position
Technology roadmap
Milestones on the path to maturity
9 milestones
Technology roadmap
Milestones on the path to maturity
Watchlists
Companies, people and papers, each with a condition for removing it
20 · 20
Watchlists
Companies, people and papers, each with a condition for removing it
Decision frameworks
What our view means for your job
Locked
Decision frameworks
What our view means for your job
Thesis changelog
When our view changed, and why
7 updates
Thesis changelog
When our view changed, and why
Change our mind
3 disconfirming conditions
You've read the verdict. The file is much deeper.
The full signal stack, technology-native KPIs tracked over time, the landscape of who depends on whom, the dated catalyst calendar, decision frameworks for every desk, live watchlists and a record of every time our view on HBM4 has changed. All of it is in CanaryIQ.