The DGX Spark 64GB goes on sale today, October 23, at a $4,999 starting price through six PC makers. It has half the memory of the original 128GB DGX Spark, yet costs $1,000 more than that machine’s $3,999 launch price. The 128GB model now lists at $6,950, nearly 75% above launch, as the memory crunch reaches desktop AI hardware.
Key takeaways
- The 64GB DGX Spark starts at $4,999, 25% above the 128GB model’s launch price.
- The 128GB DGX Spark has climbed from $3,999 to $6,950 in about a year.
- Two clustered 64GB units cost $9,998, about $3,048 more than one 128GB box.
What is the DGX Spark 64GB?
The DGX Spark 64GB is a cut-down version of Nvidia’s desktop AI computer. It keeps the same GB10 Grace Blackwell Superchip, software stack and networking as the 128GB model, but halves the unified memory. Nvidia says one unit runs models of up to 100 billion parameters, against 200 billion for the larger box.
Nvidia announced the configuration on October 2 in a blog post that set today as the on-sale date. There is no Nvidia Founders Edition this time.
Instead, the machine ships only through Acer, ASUS, Dell, Gigabyte, HP and MSI. Each partner sets its own final configuration around the storage it installs, so $4,999 is a floor, not a fixed price.
What stays the same
According to The Register, memory bandwidth holds at 273 GB/s. The CPU is still a 20-core Arm design from MediaTek, and the ConnectX-7 card still links up to four GB10 systems at 200 Gbps each.
That MediaTek link matters for investors. Nvidia’s financial ties to its Taiwanese partner run deep, as we covered in the Nvidia MediaTek bond deal.
How much does the DGX Spark 64GB cost?
The DGX Spark 64GB starts at $4,999. That is $1,000, or 25%, more than the original 128GB DGX Spark cost at its October 2025 launch. Today, the 128GB model sells for $6,950, so the smaller unit is $1,951 cheaper than its big sibling but far pricier per gigabyte of memory.
The price path tells the story. Nvidia raised the 128GB Founders Edition from $3,999 to $4,699 in February, citing “industry wide memory supply constraints” in a developer forum notice.
Then, on October 2, the 128GB model jumped again to $6,950. The Register put that at “nearly 75% above its launch price.” When Nvidia first showed the platform as Project Digits at CES 2025, the outlet notes, it was expected to cost around $3,000.
| Configuration | Date | Price | Memory | Cost per GB* |
|---|---|---|---|---|
| DGX Spark 128GB (launch) | Oct 2025 | $3,999 | 128GB | $31.24 |
| DGX Spark 128GB (first hike) | Feb 2026 | $4,699 | 128GB | $36.71 |
| DGX Spark 128GB (now) | Oct 2026 | $6,950 | 128GB | $54.30 |
| DGX Spark 64GB (new) | Oct 23, 2026 | $4,999 | 64GB | $78.11 |
| Two 64GB units, clustered | Oct 23, 2026 | $9,998 | 128GB | $78.11 |
| GMKtec Ryzen AI Max+ 395 mini PC | Oct 2026 | $3,650 | 128GB | $28.52 |
Read the last column twice. Each gigabyte of memory in the new 64GB box costs about $78, two and a half times the per-gigabyte price buyers paid a year ago. The cheaper sticker hides a worse deal.
Why is Nvidia charging more for less memory?
Memory is the expensive part now. LPDDR5x and other DRAM have soared as AI data centers absorb supply, and Nvidia has passed those costs straight to buyers. A 64GB version lets Nvidia show a sub-$5,000 price without eating the cost of 128GB of scarce memory.
This is not the first time memory has forced Nvidia’s hand. In August we reported on the 15% Nvidia AI server price hike, also driven by memory. The DGX Spark shows the same pressure moving from racks to desks.
The Register reads the unchanged 273 GB/s bandwidth as a sign Nvidia used lower-capacity LPDDR5x modules rather than fewer of them. That keeps speed intact while trimming the most expensive line on the bill of materials.
Nvidia’s pitch
Nvidia frames the 64GB unit as a way to keep the platform affordable. Its argument is that open models in the 26B-to-35B range, such as Qwen 3.8 27B, are now strong enough for local coding and research agents. Those fit comfortably in 64GB.
HotHardware put it bluntly: “It’s not half a DGX Spark; it’s a DGX Spark-like system with half the memory.” The compute is the same. The ceiling on what you can load is not.
Is clustering two 64GB units worth it?
Only if you need the bandwidth more than the savings. Two 64GB units linked by a QSFP cable pool to 128GB and run models up to 200 billion parameters, Nvidia says. But the pair costs $9,998, about 44% more than a single $6,950 128GB DGX Spark with the same total memory.
Nvidia’s case for the pair rests on one number. In its own Qwen 3.8 27B test, two clustered 64GB systems delivered “up to 1.7x the performance” of a single system, plus twice the memory bandwidth.
Be careful with that figure. Nvidia has not said whether it measures tokens per second, latency or time to first token, and no independent benchmark exists yet. A single vendor test on a 27B model that already fits in one box is thin evidence for a $10,000 purchase.
Software arriving with it
- Sync Cluster Assistant: detects connected units, validates the setup and configures ConnectX-7 networking.
- Sync Model Launcher: due at the end of October; downloads and runs Qwen3.8 27B and configures OpenCode.
- Bundled stack: Nvidia Agent Toolkit, CUDA-X AI libraries, Nemotron open models, Ollama, vLLM and PyTorch.
- Coming playbooks: serving LLMs with vLLM and connecting multiple DGX Sparks for distributed work.
Who should buy the DGX Spark 64GB?
Developers who need CUDA on a desk and plan to run models under roughly 35 billion parameters. If your work is inference on mid-size open models, the 64GB unit does the job for $1,951 less than the 128GB box. If you fine-tune or load 70B-plus models, it is the wrong machine.
HotHardware notes that models of around 80 billion parameters or more are the main casualties of the cut. The Register adds that the smaller pool leaves less room for fine-tuning.
The cheaper alternative is AMD. GMKtec’s Ryzen AI Max+ 395 mini PC listed at $3,650 with 128GB and $2,349 with 64GB when Nvidia announced the Spark, per HotHardware. The Register says GB10’s GPU has outperformed AMD’s chips on AI work in its testing, so buyers trade speed and CUDA for memory per dollar.
Who wins and who loses from the DGX Spark price shift?
Memory makers win the most. Every hike on the DGX Spark is a receipt for DRAM pricing power, which flows to Micron, SK Hynix and Samsung. Nvidia protects its margins, while independent developers and small labs absorb the cost.
Winners
- Memory suppliers. A desktop box rising 74% in a year on memory costs is the clearest retail proof of the shortage. Our Micron vs SanDisk breakdown covers how that shows up in earnings.
- Nvidia. It keeps a sub-$5,000 entry point without subsidizing memory, and pushes inventory risk onto six partners.
- AMD-based mini PC makers. A 128GB Ryzen system at $3,650 looks far better next to a $6,950 Spark than it did next to a $3,999 one.
Losers
- Independent developers and students. The “personal AI supercomputer” pitched near $3,000 now costs $5,000 for half the memory.
- Early-adopter OEMs. Partners carry the storage choices and pricing risk on a product Nvidia no longer sells directly at this tier.
One skeptical note on Nvidia’s “local AI” story. Its pitch is privacy and independence from the cloud. But at $78 per gigabyte, the economics of local inference just got worse, which quietly strengthens the case for renting cloud capacity, a market Nvidia also supplies. See how far that rental economy now stretches in our Amazon Nvidia chip leaseback story.
What to watch next
Three things will decide whether the DGX Spark 64GB is a smart buy or a stopgap: independent benchmarks, partner pricing above the $4,999 floor, and whether memory costs push the 128GB model higher again. Nvidia’s own software timeline gives the first concrete date.
- End of October 2026: Nvidia Sync Model Launcher ships, per Nvidia’s blog. It is the first test of the one-click clustering promise.
- October 2026: Windows PCs powered by Nvidia RTX Spark are slated to arrive from Acer, ASUS, Dell, HP, Lenovo, Microsoft and MSI this month, Nvidia says.
- Signal: the first independent reviews testing Nvidia’s 1.7x clustering claim, and which metric it describes.
- Signal: partner list prices for 64GB systems with real storage configurations, and any further move in the $6,950 128GB price.
FAQ
When does the DGX Spark 64GB go on sale?
Friday, October 23, 2026, through Acer, ASUS, Dell, Gigabyte, HP and MSI. Nvidia does not sell a Founders Edition of this model.
How much does the DGX Spark 64GB cost?
It starts at $4,999. Final prices vary by partner and storage configuration.
How much is the 128GB DGX Spark now?
$6,950, up from $3,999 at launch in October 2025 and $4,699 after a February 2026 increase.
What size models can the 64GB version run?
Nvidia says up to 100 billion parameters on one unit, and up to 200 billion when two units are clustered.
Is the 64GB DGX Spark slower than the 128GB model?
No. It uses the same GB10 chip and keeps 273 GB/s of memory bandwidth. Only memory capacity, and with it maximum model size, is reduced.
Why did the DGX Spark get more expensive?
Nvidia cited “industry wide memory supply constraints” when it raised the price in February. Memory prices have kept climbing as AI data centers absorb supply.
The bottom line
The DGX Spark 64GB is a price cut in name only. It costs $1,000 more than the original launched at and delivers half the memory, at about $78 per gigabyte. For developers running sub-35B models, it is a reasonable CUDA box. For everyone else, it is a $4,999 receipt for the memory shortage, and the real winners sit in Boise, Icheon and Suwon, not Santa Clara.
Free guide · 12 pages
AI Tools Every Investor Should Use in 2026
10 tools for research, stock signals, charts and crypto, what each one costs, three ready-made stacks from $0 a month and 7 copy-paste prompts for filings and earnings calls. Enter your email and we will send it to you.
Plus one email a week: AI Money This Week, every Sunday. Unsubscribe anytime.
Sources
- Nvidia: DGX Spark 64GB Gives Developers More Ways to Build and Scale Local AI (Oct 2, 2026)
- Nvidia Developer Forums: 2/23/2026 Price Change Announcement
- The Register: Nvidia debuts $4,999 DGX Spark with half the RAM and storage amid memory crunch
- HotHardware: NVIDIA DGX Spark 64 GB Reduces Memory to Stem RAM Price Increases
- Implicator: Nvidia adds 64GB DGX Spark at $4,999 as 128GB model climbs to $6,950
Wealth Engine researches and drafts with AI tools and checks every figure against the sources above. How we report.
Leave a Reply