Hardware
Chips, accelerators and the physical substrate AI runs on.
-
Tenstorrent raises $1.5bn to put its Galaxy AI servers into volume
The RISC-V AI chip company is now valued at $6.2bn and will spend the money on servers, open-weight models and chip licensing.
-
NVFP4 explained in 2 minutes
NVFP4 is a way of storing an AI model’s numbers in just 4 bits each, so big models need far less memory…
-
Inference chip
A processor designed mainly to run trained AI models and answer requests, rather than to train them.
-
NVFP4
Nvidia’s 4-bit number format for AI models, which stores each value in 4 bits with a shared scale for every 16 values.
-
TOPS
Trillions of operations per second, a headline measure of how much arithmetic an AI chip can do.
-
HBM
High-bandwidth memory: stacks of memory chips placed right beside an AI chip to feed it data very quickly.
-
NVIDIA RTX Spark
NVIDIA’s first Windows PC superchip pairs a Blackwell RTX GPU with a 20-core Grace CPU and went on sale in laptops on…
-
Nvidia adds a 120bn Nemotron and two models built to run on one machine
Nemotron 3.5 Super VL and two smaller Bolt models are out with open weights, and every performance figure in the announcement is…
-
Agility picks a Cambridge battery firm to keep Digit 5 on two shifts
Agility Robotics has named Nyobolt of Cambridge as the exclusive battery supplier for its Digit 5 humanoid, claiming 90 minutes of work…
-
TSMC posts a record quarter, with 2nm already at 8% of wafer revenue
TSMC reported third-quarter revenue of NT$1,494.24 billion and said 2nm chips already make up 8% of wafer revenue, with 79% coming from…