AI21 Labs in possible acquisition of up to $3 billion

AMD picks up ZT Systems, and Arm’s Celestial deal sees it enter the world of networking

The investment was initially announced in mid-September

There was data sovereignty, too, and some pretty big outages. Remember?

New solution is aimed at enterprises, cloud providers, and telecom operators building data centers

Offering verifies network and accelerator integrity and can spin up validated clusters “in minutes'

AWS doubles down on custom silicon and the looming memory crisis to come

Korean giant also working with SanDisk

Acquisitions and pettiness highlighted a dynamic market

NCCL support comes to GCP, alongside launch of A2UI open source project

Yet another software firm joins Nvidia, though Linux tool will remain open source, chip giant confirms

From standing start to 'on par with InfiniBand,' Spectrum-X helps Nvidia seize 11.6% of data center Ethernet switch space

Optical module makers could face component shortage as Nvidia reportedly hoards EML supply

HBM4 memory reportedly pushed back into latter half of 2026

Programming model moves from managing thousands of low-level threads to working with high-level ‘tiles of data’

Non-exclusive deal aims to accelerate chip-design workflows

Despite hobbled hardware, DeepSeek bypasses bandwidth constraints with 'sparse attention' design

AI Factories can be built using Nvidia or Trainium chips

Company also makes UltraServers with Nvidia GB300s available, and teases Trainium4

Jensen Huang confident firm can handle AI supply chain issues

It's about optionality and certainty

Demand for its Ethernet products now ‘on par’ with that of InfiniBand

New generation offers double the bandwidth compared to prior version

Neoverse integration means custom hardware from AWS, Google, and Microsoft can integrate with Nvidia GPUs