Nvidia is expanding its DGX Spark line of desktop AI computers with a new 64GB unified-memory configuration that starts at $4,999 and goes on sale on October 23, available through Acer, ASUS, Dell, Gigabyte, HP and MSI. The company will not offer a Founders Edition of the new model; it will be sold only through those manufacturer partners.
The 64GB machine keeps the same GB10 Grace Blackwell Superchip, DGX OS and full Nvidia AI software stack as the original 128GB DGX Spark, along with ConnectX-7 networking and 273 GB/s of unified memory bandwidth. Nvidia says a single 64GB unit can run AI models with up to 100 billion parameters entirely on the device, without sending data to the cloud.
Two units, one cluster
The headline feature is clustering. Using the new NVIDIA Sync Cluster Assistant, two 64GB Sparks can be linked directly over a QSFP cable — no switch required — pooling their memory into a 128GB unified pool. Nvidia says that lifts model support to as much as 200 billion parameters, and claims up to 1.7x the performance of a single system in its own internal test on the Qwen3.8 27B model. A Sync Model Launcher arriving later this month is expected to let users download and start models with a few clicks.
The price irony
The launch lands amid a sharp rise in memory prices. The original 128GB DGX Spark debuted at $3,999; Nvidia raised it to $4,699 in February 2026 citing worldwide memory supply constraints, and reports put the current price at around $6,950. That makes the new 64GB configuration — half the memory — start $1,000 higher than the 128GB machine's original launch price, and even $300 above the current Founders Edition price. Industry watchers blame soaring DRAM costs; Micron has warned supply will stay tight through 2028.
Sources: Nvidia Launches 64GB DGX Spark at $4,999 as 128GB Hits $6,950 (AI Weekly, via Tom's Hardware) | NVIDIA DGX Spark 64GB: $4,999, October 23, Clusters to 128GB (StorageReview)
The 64GB machine keeps the same GB10 Grace Blackwell Superchip, DGX OS and full Nvidia AI software stack as the original 128GB DGX Spark, along with ConnectX-7 networking and 273 GB/s of unified memory bandwidth. Nvidia says a single 64GB unit can run AI models with up to 100 billion parameters entirely on the device, without sending data to the cloud.
Two units, one cluster
The headline feature is clustering. Using the new NVIDIA Sync Cluster Assistant, two 64GB Sparks can be linked directly over a QSFP cable — no switch required — pooling their memory into a 128GB unified pool. Nvidia says that lifts model support to as much as 200 billion parameters, and claims up to 1.7x the performance of a single system in its own internal test on the Qwen3.8 27B model. A Sync Model Launcher arriving later this month is expected to let users download and start models with a few clicks.
The price irony
The launch lands amid a sharp rise in memory prices. The original 128GB DGX Spark debuted at $3,999; Nvidia raised it to $4,699 in February 2026 citing worldwide memory supply constraints, and reports put the current price at around $6,950. That makes the new 64GB configuration — half the memory — start $1,000 higher than the 128GB machine's original launch price, and even $300 above the current Founders Edition price. Industry watchers blame soaring DRAM costs; Micron has warned supply will stay tight through 2028.
Sources: Nvidia Launches 64GB DGX Spark at $4,999 as 128GB Hits $6,950 (AI Weekly, via Tom's Hardware) | NVIDIA DGX Spark 64GB: $4,999, October 23, Clusters to 128GB (StorageReview)
