US gov’t imposes restrictions on Nvidia H20s, chipmaker claims will cost it $5.5bn
Follows Nvidia's announcement it plans to produce AI supercomputers in the US
Follows Nvidia's announcement it plans to produce AI supercomputers in the US
Canadian telco also planning to expand its Kamloops site to house a supercomputer
How many components will be purely US-based is unclear
Private cloud computing company adds Nvidia GPUs to offering
Startup has raised $72.5m since founding in 2020
Additional controls were due to come into force this week
B200s now in production
Other partners include Google, Meta, and Nvidia
Chips will be deployed at the telco’s data center by summer 2025
System developed in partnership with ExxonMobil, HPE, and Nvidia
A robot that utilizes LLMs has been trialed
Fund yet to announce any investments
New launches coincide with Nvidia's GTC conference
Nvidia, AMD, Qualcomm, and Broadcom have all been approached, sources claim
CFO Marie Myers said “recent tariff announcements have created uncertainty for our industry”
Chips will be installed in phases, with 16,000 expected to be in place by end of summer 2024
Company also revealed expectation-exceeding Blackwell sales of $11bn for Q4 2025
But it has no actual commitments
Can be scaled to include 110,000 GPUs
First US university to deploy an Nvidia DGX Blackwell SuperPOD
As Trump threatens chip tariffs, and may tighten export controls
Questions remain about the validity of DeepSeek's cost claims
EdgeCloudLink data center is among the first to get Nvidia Grace Blackwell deployment
Both companies said more details about the deployment would be revealed soon