Nvidia GB300 Server Shipments Start Sept 2025 Amid Liquid Cooling Issues and Vera Rubin Plans

Nvidia GB300 server shipments begin September 2025. Learn about the smoother Blackwell Ultra transition, ongoing liquid cooling challenges
Nvidia GB300 Server Shipments Start Sept 2025 Amid Liquid Cooling Issues and Vera Rubin Plans

Nvidia GB300 Server Shipments Ready by September 2025

Some Nvidia partners have kicked off early making, but most will start sending big loads of servers using the GB300 set up in September 2025. The swap from the old GB200 line seems to be going well, though there are still bumps with the liquid cool tech for makers.

Smooth Transition to Blackwell Ultra Architecture

The move is smoother because Nvidia chose to use the old GB200 board design again. This makes making easier. But, Nvidia now sends GB300 parts in pieces, unlike before.

  • For the old GB200, Nvidia gave out a full "Bianca" board with GPU, CPU, and memory all on one board.
  • For the new GB300, parts come alone GPU on a small disk, CPU in a tiny box, and an Axiado control chip.

This lets partners pick other board parts and base memory sets from other places. Nvidia still hands out the part that connects all and the copper main board.

Manufacturing and Shipment Outlook

The makers (ODMs) are checking and making early pieces without big roadblocks. Parts are ready in time, with more making set for late 2025. The big send-out is seen to jump in the last part of the year.

Wistron, a big name in supply, thinks money made this time will not change much because of the move from GB200 to GB300.

Persistent Liquid Cooling Challenges

Even with many GB200s sent, their liquid cool tech often fails. Reports say the main weak spot is in some snap parts that leak even after stress tests. In turn, some who run data centers choose fast set up over long job life, leading to quick shutdowns and lots of tests after putting things together. This results in high service and work costs after.

Beyond GB300 Nvidia's Vera Rubin Roadmap

Nvidia plans to bring out its next AI server set, dubbed Vera Rubin, in two parts:

  • Phase One: Will see Vera CPUs and Rubin GPUs take over, but keep the current Oberon rack layout.
  • Phase Two: Will bring in a new kind of rack that uses Vera CPUs and stronger Rubin Ultra GPUs.

The new GPUs are set to need more power than the older ones, needing more from the cool tech. Fixing these cool and leak issues is key for everyone making and selling these parts.

About the author

mgtid
Owner of Technetbook | 10+ Years of Expertise in Technology | Seasoned Writer, Designer, and Programmer | Specialist in In-Depth Tech Reviews and Industry Insights | Passionate about Driving Innovation and Educating the Tech Community Technetbook

Post a Comment