NVIDIA to Launch “Blackwell Ultra” GB300 AI Servers with Liquid Cooling by Mid-2025
NVIDIA’s next-generation "Blackwell Ultra" GB300 AI servers are set to hit the market by mid-2025, boasting substantial performance enhancements over current offerings. This marks a significant step in NVIDIA’s strategy to solidify its dominance in the AI sector, leveraging groundbreaking technologies to meet rising market demand.
The GB300 AI servers are expected to surpass the current GB200 NVL72 models in both performance and power efficiency. Here are the key highlights:
Fully-Liquid Cooling Solution: To handle the anticipated increase in power consumption, the GB300 servers will integrate a fully-liquid cooling system, ensuring optimal thermal performance. This will drive demand for cooling components, with Taiwanese manufacturers such as Auras Tech and Asia Vital Components benefiting from the shift.
Socketed GPU Design: Unlike their soldered predecessors, the GB300 GPUs will adopt a socketed design, enabling easier installation and removal, simplifying production processes, and boosting efficiency for server manufacturers.
Higher Price Points: With advanced features and fully-liquid cooling, the pricing of the Blackwell Ultra GB300 servers is expected to exceed the $3 million cost of current GB200 configurations, catering to high-end enterprise and research clients.
NVIDIA's Blackwell Ultra servers are poised to capitalize on the surging demand for AI solutions, driven by tech giants integrating AI into their operations. Taiwanese suppliers are already preparing for increased demand for components, with liquid cooling and interconnect technologies playing pivotal roles in the GB300 rollout.
If successful, the launch will further boost NVIDIA’s revenue and market leadership, reinforcing its position at the forefront of AI hardware innovation.
The Blackwell Ultra GB300 series is expected to debut mid-2025, setting new benchmarks for AI server performance. As NVIDIA transitions to this advanced platform, it aims to address the needs of AI-driven workloads with unparalleled efficiency and scalability.
What are your thoughts on NVIDIA’s ambitious plans for the GB300 AI servers? Will fully-liquid cooling and socketed designs reshape the AI server market? Share your views below!
