Tap a claim on the ring to put it at the centre.
← The rush to build high-bandwidth memory starves ordinary DRAM and NAND…
8 connected korrents · 5 moments on record from 3 Mar 2024 to 18 Jun 2026.
−
+
Fit
Full text
Drag to move · pinch or scroll to zoom
Everything filed under semiconductors
semiconductors Same subject Same subject Same subject Same subject Same subject Same subject Same subject Same subject Read this korrent: The rush to build high-bandwidth memory starves ordinary DRAM and NAND of capacity, so AI demand tightens the memory markets it never touches. The rush to build high-bandwidth memory starves ordinary DRAM and NAND of capacity, so AI demand tightens the memory markets it never touches. Last stated 3 years ago 3 Mar 2024 · 3 years ago
DO
Doug O'Laughlin — holds since 2024-03-03 — tap for who they are
Same subject: The faster AI progresses, the better NVIDIA does, so a shock like DeepSeek should expand its market rather than shrink it. — tap to centre the map on it The faster AI progresses, the better NVIDIA does, so a shock like DeepSeek should expand its market rather than shrink… Last stated 2 years ago 3 Feb 2025 · 2 years ago
NL
Nathan Lambert — holds since 2025-02-03 — tap for who they are
Same subject: The binding constraints on AI's growth are power, helium and memory — and almost nobody has noticed the helium one. — tap to centre the map on it The binding constraints on AI's growth are power, helium and memory — and almost nobody has noticed the helium one. Last stated 3 months ago 18 Jun 2026 · 3 months ago
LT
Lip-Bu Tan — holds since 2026-06-18 — tap for who they are
Same subject: The memory crunch will make ordinary phones and computers worse year on year, and that is why the public will come to hate AI more. — tap to centre the map on it The memory crunch will make ordinary phones and computers worse year on year, and that is why the public will come to… Last stated 6 months ago 13 Mar 2026 · 6 months ago
DP
Dylan Patel — holds since 2026-03-13 — tap for who they are
Same subject: High-bandwidth memory costs roughly twice the capital per bit that ordinary DRAM does, because its dies are twice the size. — tap to centre the map on it High-bandwidth memory costs roughly twice the capital per bit that ordinary DRAM does, because its dies are twice the… Last stated 3 years ago 3 Mar 2024 · 3 years ago
DO
Doug O'Laughlin — holds since 2024-03-03 — tap for who they are
Same subject: The reason we cannot simply make more memory is that there is nowhere to put the tools, and a fab takes two years to build. — tap to centre the map on it The reason we cannot simply make more memory is that there is nowhere to put the tools, and a fab takes two years to… Last stated 6 months ago 13 Mar 2026 · 6 months ago
DP
Dylan Patel — holds since 2026-03-13 — tap for who they are
Same subject: What an accelerator needs from memory is bandwidth per unit of chip edge, and on that measure ordinary DRAM is an order of magnitude behind HBM. — tap to centre the map on it What an accelerator needs from memory is bandwidth per unit of chip edge, and on that measure ordinary DRAM is an… Last stated 6 months ago 13 Mar 2026 · 6 months ago
DP
Dylan Patel — holds since 2026-03-13 — tap for who they are
Same subject: Falling back to older 7-nanometre chips would not rescue AI compute, because the real gap between chip generations is twentyfold, not the threefold the flops suggest. — tap to centre the map on it Falling back to older 7-nanometre chips would not rescue AI compute, because the real gap between chip… Last stated 6 months ago 13 Mar 2026 · 6 months ago
DP
Dylan Patel — holds since 2026-03-13 — tap for who they are
Same subject: There is no AI bubble on the demand side: the shortage of computing capacity is global, and spans every company and industry. — tap to centre the map on it There is no AI bubble on the demand side: the shortage of computing capacity is global, and spans every company and… Last stated 8 months ago 8 Jan 2026 · 8 months ago
JH
Jensen Huang — holds since 2026-01-08 — tap for who they are
At the centre
The rush to build high-bandwidth memory starves ordinary DRAM and NAND of capacity, so AI demand tightens the memory markets it never touches.
Holds DO Doug O'LaughlinRead this korrent →
The faster AI progresses, the better NVIDIA does, so a shock like DeepSeek should expand its market rather than shrink it.
Holds NL Nathan Lambert
The binding constraints on AI's growth are power, helium and memory — and almost nobody has noticed the helium one.
Holds LT Lip-Bu Tan
The memory crunch will make ordinary phones and computers worse year on year, and that is why the public will come to hate AI more.
Holds DP Dylan Patel
High-bandwidth memory costs roughly twice the capital per bit that ordinary DRAM does, because its dies are twice the size.
Holds DO Doug O'Laughlin
The reason we cannot simply make more memory is that there is nowhere to put the tools, and a fab takes two years to build.
Holds DP Dylan Patel
What an accelerator needs from memory is bandwidth per unit of chip edge, and on that measure ordinary DRAM is an order of magnitude behind HBM.
Holds DP Dylan Patel
Falling back to older 7-nanometre chips would not rescue AI compute, because the real gap between chip generations is twentyfold, not the threefold the flops suggest.
Holds DP Dylan Patel
There is no AI bubble on the demand side: the shortage of computing capacity is global, and spans every company and industry.
Holds JH Jensen Huang