NVIDIA's new GB300 NVL72 AI server in the flesh at Computex 2025, packing Blackwell Ultra GPUs

As an Amazon Associate, we earn from qualifying purchases. TweakTown may also earn commissions from other affiliate partners at no extra cost to you.

NVIDIA's beasty new GB300 NVL72 AI server racks are on show at Computex 2025, where an early walk around the halls of Nangang seeing me instantly attracted me... nerd vibes were at 11/10. Check it out:

6

ASRock had the new NVIDIA GB300 NVL72 AI server rack at their booth, with the fully liquid-cooled, rack-scale system combining 72 x GB300 "Blackwell Ultra" AI GPUs with an upgraded 288GB of HBM3E memory, 1.4kW of power, and 50% more performance than GB200 NVL72 AI servers.

6

NVIDIA is expected to have its new GB300 NVL72 AI servers enter mass production in Q3 2025, according to recent supply chain sources. Apple reportedly placed an order for around $1 billion worth of new NVIDIA GB300 AI chips, with each GB300 NVL72 AI server costing between $3.7 and $4 million a pop... sadly, it wouldn't fit in my backpack so I left it on the trade show floor.

• Read more: NVIDIA's next-gen GB300 AI server mass production might not happen until 2026
• Read more: Apple becomes NVIDIA customer, orders estimated $1 billion of GB300 NVL72 AI servers
• Read more: NVIDIA GB300 'Blackwell Ultra' AI GPU: 288GB HBM3E, 50% faster than GB200

NVIDIA's new GB300 NVL72 AI server connects 72 x GB300 Blackwell Ultra AI GPUs and 36 Arm Neoverse-based NVIDIA Grace CPUs in a rack-scale design, acting as a single gigantic GPU built for test-time scaling. NVIDIA's new GB300 NVL72 has AI models accessing the platform's massive performance gains to explore different solutions to problems and break down complex requests into multiple steps, resulting in higher-quality responses.

6

NVIDIA explains its new GB300 "Blackwell Ultra" AI platform: "NVIDIA Blackwell Ultra boosts training and test-time scaling inference - the art of applying more compute during inference to improve accuracy - to enable organizations everywhere to accelerate applications such as AI reasoning, agentic AI and physical AI".