AI Workstation Nvidia DGX Spark: Price increase due to higher memory costs
AI boom makes memory expensive. Nvidia, a key driver, raises prices for its DGX Spark AI mini-workstation due to higher memory costs.
(Image: Nvidia)
It's a vicious cycle, as usual, that customers have to bear the brunt of. One of the main beneficiaries of the AI boom is increasing the prices of its AI mini-workstation from around $4,000 to $4,700, an increase of about 17.5 percent. We're talking about Nvidia, which recently raised its non-binding price recommendation for the DGX Spark AI mini-workstation. As a reason, they cited in their own developer forum “Memory Supply Constraints,” meaning the currently extremely tight memory supply.
That this affects the AI beneficiary Nvidia in particular could be considered a fine irony. However, at least the increased procurement costs will be passed on directly to the customers, even if the target group is not the typical office user or gamer, but AI developers.
The DGX Spark also has 128 GB of LPDDR5X RAM soldered on. The memory, which the CUDA cores of the GB10 chip have equal access to, delivers and receives data at 273 gigabytes per second. Also in the blast radius of the memory price: the SSD, which holds 4 TB at Nvidia. Comparable offers from other providers such as Acer, Asus, Dell, Gigabyte (to the c't test), HP or Lenovo are usually more spartan and start with a single terabyte of SSD storage.
(Image: Nvidia, Screenshot: c’t)
The amount of RAM, on the other hand, is fixed at 128 GB, as is the other hardware equipment with the 200 Gbit/s network adapter ConnectX-7 from Nvidia's Mellanox division, a 10 Gbit/s Ethernet port, Wi-Fi 7, and four USB-C sockets. One of which is for power supply.
Videos by heise
Overplayed its hand?
This price increase is somewhat surprising. Because just a few weeks ago, Nvidia CEO Jensen Huang said that Nvidia had made provisions for AI accelerators. However, this was probably not mainly about standard chips like GDDR7, DDR5, and LPDDR5X, but primarily about the super-expensive High Bandwidth Memory of the third and fourth generations (HBM3e/HBM4), where there is also significantly less competitive pressure. And for the technically complex HBM stacks, Nvidia enters into long-term contracts and even development partnerships anyway.
Empfohlener redaktioneller Inhalt
Mit Ihrer Zustimmung wird hier ein externer Preisvergleich (heise Preisvergleich) geladen.
Ich bin damit einverstanden, dass mir externe Inhalte angezeigt werden. Damit können personenbezogene Daten an Drittplattformen (heise Preisvergleich) übermittelt werden. Mehr dazu in unserer Datenschutzerklärung.
In any case, the equipment of the DGX Spark was probably not part of Nvidia's provisions. However, Nvidia also states in its statement that pre-orders will be delivered at the prices valid at the time of contract conclusion. They have no influence on the pricing of other providers, some of whom entered the market at prices up to 1000 euros below Nvidia's DGX Spark.
(csp)
