MicroLEDs Ain’t Dead, Micron Snags Vera Rubin

20 snips

Mar 20, 2026

They debate Jensen Huang's $250K-per-engineer AI token claim and whether companies will buy tokens or on-prem hardware. Micron's blockbuster earnings and a Vera Rubin design-in for HBM4 drive a deep dive into pin speeds, base-die tradeoffs, and big new fab bets. The conversation ends on optical interconnects, where microLEDs, VCSELs, and a new OCI-MSA standard vie as short-reach solutions.

Ask episode

AI Snips

Chapters

Transcript

Episode notes

ADVICE

Buy Tokens Or Token Generators

Consider buying on-prem token generators instead of unlimited cloud token budgets.
Austin suggests teams could buy Dell Pro Max–class machines (~$120k) to run inference locally and avoid $/token OPEX unpredictability.

INSIGHT

Tokens Will Be Shared Like EDA Licenses

Shared on-prem GPU capacity will likely behave like EDA license pools with contention controls.
Vikram compares token/compute clusters to LSF clusters where users submit jobs and wait for resources to free up.

ANECDOTE

Running OpenClaw On A Home Server

Vikram deployed OpenClaw on a home server inside a Docker container with a local reverse proxy and TailScale VPN.
He fed it Claude credits and keeps the agent local to preserve memories and reduce token cost.

Get the Snipd Podcast app to discover more snips from this episode

Get the app

Austin and Vik break down a packed week in semiconductors, covering GTC, OFC, and Micron earnings. The conversation kicks off with Jensen Huang's bold claim that engineers should spend $250K/year on AI tokens, and whether companies will buy tokens or token generators (i.e., on-prem hardware like the Dell Pro Max with GB300). They dig into the CapEx vs OpEx tradeoffs, data security concerns, and how sharing GPU resources might end up looking a lot like the old EDA license model.

Next up: Micron crushed earnings and appears to be designed into Vera Rubin for HBM4 — despite months of rumors saying otherwise. Austin and Vik unpack the nuance around HBM pin speeds, memory node base dies, and what Micron's massive new fab investments in Taiwan, Singapore, Idaho, and New York mean for the memory cycle.

The back half of the episode dives into optical interconnects for AI scale-up. A new industry consortium (OCI-MSA) has formed with Meta, Broadcom, NVIDIA, and OpenAI to standardize optical components. Vik explains why traditional indium phosphide lasers might be overkill for short-reach scale-up, and makes the case for micro LEDs — a "slow but wide" approach that could fill the gap between copper and conventional optics. They also touch on Credo's expanding product portfolio (and the infamous purple-to-orange cable saga), plus Lumentum's new VCSEL work for scale-up.

Vik - https://www.viksnewsletter.com/
Austin - https://www.chipstrat.com/

CHAPTERS
0:00 Intro & GTC/OFC Conference Overload
2:09 Jensen's $250K Token Budget Per Engineer
5:08 On-Prem Inference vs. Cloud Token Spending (Dell Pro Max, CapEx vs OpEx)
6:44 Sharing GPU Resources Like EDA Licenses
8:16 Data Security & On-Prem Privacy Concerns
9:53 Matthew Berman's Fine-Tuned Open Claw Agent
10:35 Vik Sets Up Open Claw on a Home Server
11:53 Always Be Clauden (ABC) – Managing Agents from Your Phone
13:34 Micron Earnings & HBM4 in Vera Rubin
16:39 HBM Pin Speeds & the Micron Design-In Debate
20:17 Micron's New Fab Investments & Memory Cycle Fears
23:49 Why AI Drives a Step Change in Memory Demand
26:30 Optical Compute Interconnect MSA (OCI-MSA)
29:48 Scale-Up Optics: Do We Need New Technology?
30:58 Micro LEDs – The "Slow but Wide" Approach
35:45 Micro LEDs vs. Copper vs. Traditional Optics
36:55 Credo's Product Spectrum & the Purple Cable Story
39:31 VCSELs & Lumentum's 1060nm Scale-Up Play