Product & chips · Jul 25, 2026

The MI455X physical: process and packaging fully ahead, the "brain" a generation behind.

Chips & semiconductors · This week (published July 25)

From the Jul 25, 2026 daily brief

AMD's next flagship AI chip, the MI455X, is the first datacenter chip on TSMC's 2nm process; its 3,470 mm² of logic silicon is the most in any single package in the industry, and its 12-stack HBM4 memory totals 432GB against 288GB for Nvidia and Google — all hard leads. But SemiAnalysis rates the GPU's microarchitecture a generation behind Nvidia and converging toward its rival in multiple places (down to native support for Nvidia's 4-bit number format) — silicon area brute-forcing a design gap. Microarchitecture is the chip's internal design itself: it decides how much performance the same silicon can squeeze out. Nvidia, in response, is pushing its next-generation memory pin speeds beyond the official spec to claw back the bandwidth gap (SemiAnalysis, Jul 25). Same source and article as thread 1: the specs have official grounding, the ratings are one firm's judgment.

More in this section

Product & chips Sep 10, 2026 1 sources
Customers of chip-design tool vendors have started capping tokens engineer by engineer — and a year ago nobody cared what that line cost.
Product leaders from five electronic design automation and AI chip-design tool vendors sai…
Product & chips Sep 10, 2026 1 sources
The new flagship's enterprise edition went live with its first published prices, and its own account of the mechanism doubles as an audit cost.
OpenAI's enterprise announcement on September 9 recast this model from "the smartest model…
Product & chips Sep 9, 2026 1 sources
OpenAI says its first in-house inference chip delivers 1.5 to 1.9 times the throughput per watt of "the commercial systems tested" — and does not say which systems those were.
A document signed by CFO Sarah Friar gives the first numbers for Jalapeño, OpenAI's first …
Product & chips Sep 9, 2026 1 sources
Meta launched a personal AI agent called Muse, and the part worth recording is not what it does but that a payment network has written terms for a non-human buyer for the first time.
Meta introduced Muse on September 8: it runs on a cloud virtual machine dedicated to each …

The MI455X physical: process and packaging fully ahead, the "brain" a generation behind.

Subscribe free — first issue lands tomorrow morning

More in this section

Customers of chip-design tool vendors have started capping tokens engineer by engineer — and a year ago nobody cared what that line cost.

The new flagship's enterprise edition went live with its first published prices, and its own account of the mechanism doubles as an audit cost.

OpenAI says its first in-house inference chip delivers 1.5 to 1.9 times the throughput per watt of "the commercial systems tested" — and does not say which systems those were.

Meta launched a personal AI agent called Muse, and the part worth recording is not what it does but that a payment network has written terms for a non-human buyer for the first time.