All posts in this stream, newest first.
Nvidia's new prefabricated DSX POD promises to cut AI data center build times from months to weeks, but a Greenpeace protest at the Berlin unveiling highlighted growing concern over the industry's environmental impact.
Cisco, Broadcom, and Nvidia are racing to ship high-buffer, high-bandwidth switching silicon that keeps GPU clusters from becoming expensive, underutilized hardware. Hyperscalers are expected to adopt first, with enterprise trickle-down to follow.
The AI inference provider is turning to a competing cloud giant for a large cluster of Nvidia's HGX B300 systems, underscoring just how tight AI infrastructure supply remains.
