Samsung and SK Hynix each committing ₩10-20 trillion for ASML EUV tools as ASML warns of 2027-2028 supply shortfall
FollowingDay 2in the edition since 9 Nov 20263 sourcesupdated 9 Nov 2026
Samsung Taylor Fab 2 and AMD HBM
The gist
Latest: AMD reportedly secures most of Samsung 2nm Taylor Fab 2 output for HBM
The reported capacity pre-booking commits a large share of Samsung's advanced foundry node to AMD before the fab is even built, years before production. The split, AMD at Samsung and Nvidia at TSMC, shapes which memory makers and foundries serve the two largest AI chip vendors.
When stories like this move, I write about them in an email each morning.
One free email each morning, around 8 AM US Eastern.
Check your email.
I have sent you a confirmation link. One click and you are in.
Not in your inbox? Look in spam or promotions. If it's there, mark it "Not spam" so future editions land in your inbox.
The full picture
Samsung's Taylor Fab 2 has reportedly pre-booked its entire planned 2nm HBM base-die output before construction begins, with AMD taking most of the secured capacity. Equipment installation at the Texas facility is expected around 2029 and mass production in 2030. Samsung has not officially confirmed which products the fab will manufacture. Nvidia is pursuing a different arrangement, reportedly asking Samsung, SK Hynix, and Micron to build its custom NVHBM base dies on TSMC's 3nm process. That structure would generate memory-related revenue for TSMC without requiring it to enter the DRAM market directly. TSMC denied a separate report that Nvidia had pushed it toward 3D DRAM, stating it will remain a pure logic foundry.
How it developed
First in the edition · 9 November 2026
AMD reportedly to use Samsung 2nm at Taylor for HBM base dies; Nvidia still prefers TSMC
Tweaktown and igorslab corroborated and extended earlier reports that Samsung's planned Taylor Fab 2, a 250,000-square-meter plant using a 2nm-class process, has secured orders covering its entire planned output for HBM base dies before construction begins, with AMD accounting for most of the volume.
Equipment installation is expected in 2029 and mass production in 2030. Nvidia is also named as a potential buyer of base dies from the same facility.
Nvidia and Google reported co-designing custom HBM base dies with memory makers; Korean media expects custom HBM market to open fully in 2026
Sources
Related
- Related storyNvidia Rubin memory cuts and HBM shortage
- Nvidia's NVHBM base die at TSMC
- Nvidia AI server prices and memory shortage
- Hyperscaler Q3 earnings and AI supply crunch
- Anthropic's custom silicon team for Claude
BackgroundCompute: the chips, clouds and data centers behind AI
Get the daily email
When stories like this move, I write about them in an email each morning.
One free email each morning, around 8 AM US Eastern.
Check your email.
I have sent you a confirmation link. One click and you are in.
Not in your inbox? Look in spam or promotions. If it's there, mark it "Not spam" so future editions land in your inbox.