Rubin makes inference a rack problem
The emphasis turns to memory, power, and racks as interconnected parts of the puzzle
Tensordyne bets on math to beat Nvidia
Startup relies on logartihmic math to replace multiplication functions with addition wherever possible, and aims to cut power and gain room for memory.
Nvidia shrinks Blackwell for mass robotics
New modules target mainstream edge AI
IBM’s latest AI processor is GPU-free
It does have 32 NPUs, however.
Sunrise releases next-generation inference GPU, Qiwang S3
Is shifting its architecture toward inference-first design.
The proof is in the shipping
Furiosa sends 4,000 AIP AIBs off to work.
Nvidia makes a dramatic, gigantic acquisition-licensing deal with Groq
As JPR predicted, the biggest private companies will get acquired.
Vsora shifts from vehicles to data center AI inferencing
Chiplet design taped-out, production 2026.
Intel Unveils Crescent Island GPU for the AI inference era
The workload is shifting to smart devices
Meta swallows Rivos—now what?
Lots of hints, rumors, and leaks to suggest Meta’s ambitions for Rivos may go beyond adding to Meta’s technology portfolio.
Meta’s next-generation MITA Training and Inference AIP
MTIA complements and strengthens Meta’s AI infrastructure investments.
Nvidia Rubin CPX and disaggregated long-context inference
Massive context and the inference dichotomy.