Nvidia’s 64GB DGX Spark adds a lower-capacity option to its compact AI workstation lineup, but it is not cheaper than the original Spark was at launch. Scheduled for October 23, 2026, at a starting price of $4,999, it has half the memory and storage of the original configuration. AMD’s $3,999 Ryzen AI Halo offers another option, with 128GB of unified memory.
3
4
12
23
DGX Spark 64GB: specs, price and availability
The new configuration keeps the GB10 Grace Blackwell platform and pairs it with 64GB of unified LPDDR5X memory and 2TB of storage. The original Spark configuration has 128GB of memory and 4TB of storage.
2
3
12
The 64GB version is scheduled to go on sale October 23, starting at $4,999. It will be sold through Acer, ASUS, Dell, Gigabyte, HP and MSI; Nvidia is not offering a Founders Edition in this configuration.
3
8
10
Nvidia says a single 64GB Spark can run models with up to 100 billion parameters locally. Treat that as a stated capability, not a promise that every model of that size will fit or run at a useful speed: the result depends on the model and workload. Two Spark units can also be linked using ConnectX-7 networking.
3
12
How it compares with the original 128GB Spark
The main trade-off is capacity versus upfront cost. The 64GB version has half the original model’s memory and storage, but it costs $4,999—$1,000 more than the original 128GB Spark’s $3,999 launch price. Meanwhile, the 128GB model is now reported at $6,950, making the new version $1,951 less expensive than that current reported price.
3
4
5
The 128GB model’s reported price is nearly 75% above its launch price. Nvidia had already raised it to $4,699 in February 2026; at the time, the company cited memory supply constraints.
2
40
41
AMD’s Ryzen AI Halo alternative
AMD’s Ryzen AI Halo is priced at $3,999 and is advertised with 128GB of unified memory—twice the capacity of the 64GB Spark for a $1,000 lower starting price. AMD says the Halo can run models up to 200 billion parameters with the right quantization, but that is also a vendor capability claim, not a guarantee for every model or use case.
23
29
That comparison is about stated configurations and starting prices, not a full performance verdict. Buyers should check the exact system, software support and workload requirements before choosing; the supplied sources do not provide a like-for-like benchmark of the two machines.
What memory shortages mean for prices
The reported price changes show why comparing the new Spark only with the original launch price can be misleading: its $4,999 entry price is lower than today’s reported $6,950 price for the 128GB version, but higher than the 128GB model’s $3,999 launch price. Reporting also links the later price increase to tight memory supplies, following Nvidia’s February move, which the company attributed to memory constraints.
2
3
40
41
That evidence supports caution, not a precise forecast. The sources do not establish when memory supply will improve or how that would affect the prices of DGX Spark systems or competing hardware.
Bottom line
The 64GB DGX Spark lowers the current cost of entry into Nvidia’s compact workstation lineup, while cutting memory and storage in half compared with the original configuration. AMD’s Ryzen AI Halo has a lower stated starting price and 128GB of memory, but the available sources do not settle which system performs better for a particular task. Compare the exact configuration and workload—not parameter-count claims alone.