Friday, February 27th 2026

NVIDIA Raises DGX Spark Pricing to $4,700

NVIDIA is raising the MSRP of the DGX Spark AI development standalone to $4,700, which represents a $700 increase from its launch price, or a 17.5% increase. This increase in pricing adjusts for industry-wide shortage of DRAM and NAND flash, which impact the prices of the 128 GB LPDDR5X unified memory the machine uses. "The MSRP for DGX Spark (Founders Edition) has been adjusted from $3,999 to $4,699 due to memory supply constraints," NVIDIA said in a statement on its Developer forums. The new price is set to go into effect later this week. This should also impact prices of custom-design variants of the DGX Spark released by various partners.
Sources: NVIDIA Developer Forums, VideoCardz
Add your own comment

5 Comments on NVIDIA Raises DGX Spark Pricing to $4,700

#2
framebuffer
Is anybody except YouTubers even buying this things?
Here in Switzerland I saw the ASUS one as low as CHF 2300 (~$3K), currently it's up a bit but still far from those $4.7K
Posted on Reply
#3
Space Lynx
Astronaut
such is the nature of shortages, not really a big deal
Posted on Reply
#4
igormp
framebufferIs anybody except YouTubers even buying this things?
Here in Switzerland I saw the ASUS one as low as CHF 2300 (~$3K), currently it's up a bit but still far from those $4.7K
I think this is pretty much similar to the GeForce FE offerings, more of a reference thing whereas most people end up with products from other OEMs.
In this specific case, Nvidia's first party offering is way more expensive than the ones from Asus, Gigabyte, HP or Dell.
Posted on Reply
#5
Buddyspike
framebufferIs anybody except YouTubers even buying this things?
Here in Switzerland I saw the ASUS one as low as CHF 2300 (~$3K), currently it's up a bit but still far from those $4.7K
People who don’t know( or care that much ) about the price difference for essentially same product (in terms of internals). Also I believe the 3000€ Asus is only 1TB instead of 4TB in DGX Spark.
But yeah , I read people buy them rather for experimentation with LLMs , then anything close for being enough for production use ( due to performance) . It has decent amount of unified RAM , but that’s about it.
even in a small company, for local LLMs you’re much better off with multiple workstation GPUs or at least something like a single RTX Pro 6000 (or dual 5090 or 4090s etc. ).
Of course buying online subscriptions is usually still way cheaper option, but has some downsides.
Posted on Reply
Nov 22nd, 2026 11:55 CST change timezone

New Forum Posts

Popular Reviews

TPU on YouTube

Controversial News Posts