Nvidia Prices VR200 NVL72 at $7.8 Million as Memory Costs Jump 435% to $2 Million
Updated
Updated · Tom's Hardware · May 21
Nvidia Prices VR200 NVL72 at $7.8 Million as Memory Costs Jump 435% to $2 Million
2 articles · Updated · Tom's Hardware · May 21
$7.8 million is Morgan Stanley’s estimate for each Nvidia VR200 NVL72 rack sold to hyperscalers, nearly double the roughly $4 million price of a GB300 NVL72 system.
$2 million of that total comes from memory, now about 25% of the rack’s cost after a 435% jump driven by far higher LPDDR5X and 3D NAND content.
54 TB of LPDDR5X goes into each VR200 NVL72 versus 17 TB in GB200 NVL72, while 3D NAND storage rises to about $1 million from virtually zero in the earlier system.
Nvidia is also lifting the rest of the bill of materials with Rubin GPUs priced at $55,000, Vera CPUs at $5,000, and more complex switching, networking, cooling, power and packaging.
The estimates underscore how AI-server demand is pushing memory and infrastructure costs sharply higher for cloud providers building next-generation Nvidia clusters.
With AI racks nearing $8 million, is Nvidia's pricing creating an opening for its rivals?
As one AI rack consumes more power than 150 homes, is the energy cost of AI becoming unsustainable?
Is Nvidia's control over system design turning the world's biggest tech companies into mere assemblers?
Nvidia’s Vera Rubin NVL72 Sets a $7.8 Million Standard: Unpacking the Cost Surge and Supply Chain Crisis in AI Supercomputers
Overview
Nvidia's Vera Rubin NVL72 rack sets a new standard in AI infrastructure with its $7.8 million price, which is double that of the previous Blackwell generation. This dramatic price increase marks a major shift in the economics of deploying advanced AI, as nearly all components—not just memory—have become more expensive. The surge reflects a broad rise in the cost of sophisticated hardware needed for top-tier AI computing. As a result, organizations must now make much larger investments to stay at the forefront of AI development, highlighting the growing challenges and changing landscape of high-end AI technology.