NVIDIA DGX Spark 64GB Launched and Big 128GB GB10 Price Increases
NVIDIA has introduced a new 64GB version of its DGX Spark AI system at $4,999 while simultaneously increasing the price of its 128GB model to $6,950. The company appears to be addressing market demand with the more affordable 64GB option as memory costs continue to rise across the industry.
NVIDIA is making significant changes to its GB10 ecosystem, introducing a more affordable 64GB model while increasing prices for the higher-end 128GB version. The company appears to be responding to market demand with the new entry-level option, which retains the NVIDIA ConnectX-7 200GbE networking found in the 128GB model.
The price increase for the 128GB model is substantial, jumping from $3,999 at launch to $6,950 now. This significant hike has made the system less accessible for many customers, prompting NVIDIA to develop the 64GB alternative as a more budget-friendly option.
New 64GB DGX Spark Available from Major OEMs
NVIDIA announced that the new DGX Spark 64GB will be available from major OEMs including Acer, ASUS, Dell, Gigabyte, HP, and MSI starting Friday, Oct. 23, with a price point of $4,999. The company has not yet clarified whether this pricing refers to the 1TB or 4TB configuration of the system.
According to NVIDIA, customers can easily cluster two new 64GB models for 128GB capacity at around $8,000, with software support for this configuration. John Lee noted that when building an 8x NVIDIA GB10 128GB cluster earlier in 2026, nodes were purchased for around $3,000 each, making it now a $50K+ option.
128GB Model Price Increases to $6,950
The price increase for the 128GB DGX Spark represents a significant jump from its original $3,999 launch price to $6,950. NVIDIA has confirmed the new price point for this model, which now sits at approximately $6,950 depending on the OEM.
John Lee commented that while $4,999 for a 64GB machine and $6,950 for a 128GB machine may feel high, these prices put the 128GB model in competition with AMD's Gorgon Halo systems offering 192GB of memory capacity. The choice becomes whether to prefer AMD with 50% more memory capacity or NVIDIA with better networking but lower capacity.
Market Context and Competitors
Memory prices continue to increase largely due to models improving, which raises the value proposition of the GB10 despite higher hardware costs. Lee noted that $6,950 is almost equivalent to three years of a $200/month OpenAI or Claude subscription, though cloud solutions involve sending data to third parties.
Patrick Kennedy, EiC and Chief Analyst, explained that memory costs justify the price increases, stating they are paying $2,000 for 64GB ECC RDIMMs. He also acknowledged that the team underestimated how quickly they could generate value from the GB10 systems, getting more out of them than projected.
We were far off in how fast we have been able to generate value from the GB10's as we have been getting much more out of them than we projected.
Apple's upcoming $10,000 Mac Studio M5 Ultra with 256GB memory capacity will emerge as a strong competitor if two NVIDIA GB10s cost more. Lee mentioned their lab's M3 Ultra 512GB unit has been an absolute workhorse, suggesting Apple's high-end systems offer significant competition in this space.
