Direct answer: When we last read NVIDIA Marketplace on October 6, the 128 GB / 4 TB DGX Spark was listed at $6,950 and out of stock. The Marketplace page did not load for us on October 23, so we cannot tell you today’s stock. If you need a Spark-class system now, you have four realistic options: wait for restock, buy a partner’s GB10 system, rent a cloud GPU while you wait, or pay for a turnkey deployment. Aradia’s listed turnkey Spark is $15,125 with hardware included and a stated 30-day deployment timeline. That is more than twice the last bare-hardware price we saw, and it comes with a cancellation fee. This page lays out when each option makes sense.
Stock and prices move quickly, so treat every figure here as a dated observation, not a quote.
First, check whether the shortage is still real
Before changing your plan, check three places on the same day: NVIDIA Marketplace for the 128 GB unit, the partner stores for the systems in our OEM comparison, and any reseller you trust. On October 23, NVIDIA’s product page showed a “Buy Now” button that points to the Marketplace. The “64 GB Coming Soon” notice we saw on October 7 was no longer in the page text we read. We did not verify actual partner listings or stock, so do not take that as proof that anything is available.
Option 1: Wait for the NVIDIA listing
Waiting is the cheapest path if your project can slip. The last observed price was $6,950, but our records also show $4,699 on September 11 and no displayed price on September 29, so the number is not stable. Waiting makes sense when you only need a development box and a few weeks do not hurt. It does not make sense if a customer deadline, a hiring plan or a budget cycle depends on having the system running.
Option 2: Buy a partner GB10 system
Acer, ASUS, Dell, HP and Lenovo sell systems on the same GB10 platform, and NVIDIA names Gigabyte and MSI as partners as well. Our OEM comparison lists what each vendor’s own pages showed. Dell’s selected 128 GB / 4 TB configuration showed $9,007 in our last check, which is well above the NVIDIA listing price, so compare configurations carefully. NVIDIA’s October 2 announcement said 64 GB versions would start at $4,999 from partners, but a 64 GB system holds smaller models; see our 64 GB versus 128 GB guide before choosing it just for the price.
The catch: partner systems are hardware only. You still have to install, configure and operate the software stack.
Option 3: Rent a cloud GPU while you wait
If your real goal is to start experimenting, you may not need to own hardware yet. Renting a GPU lets you test your models and workloads this week and gives you real numbers before you spend thousands of dollars. Our guide to renting GPU time while the Spark is out of stock explains how to start and what to check, and our DGX Spark versus cloud GPUs comparison and cloud GPU cost calculator show how to estimate when buying becomes cheaper than renting, and you can try your own numbers in our interactive break-even calculator. Cloud rental is a poor fit when your data cannot leave your premises or when you will run the system around the clock for years.
Option 4: Turnkey deployment
Aradia sells a DGX Spark as a finished private AI system. On October 23, its pricing page listed:
- $15,125 for the turnkey DGX Spark, with the hardware included (128 GB memory and 4 TB storage listed);
- a 30-day deployment timeline;
- an optional monthly SLA of $1,500;
- a 12-month hardware warranty and a 14-day re-staging guarantee if configuration drift is detected;
- change orders outside the included SLA hours billed at $450 per hour.
Two things deserve attention. First, the price gap to bare hardware is not a hidden fee: it pays for staging, model configuration, benchmarking and hardening that you would otherwise do yourself. Our bare versus turnkey guide breaks that down. Second, Aradia’s page says that if you cancel before deployment, it keeps a non-refundable fee of 40% of the Spark turnkey total. On a $15,125 order, that is about $6,050, so do not sign until the scope and delivery date are in writing.
We could not confirm from the pricing page whether Aradia has Spark units in stock today. Ask for a written delivery commitment before you pay anything.
Which option fits you?
| Your situation | Likely best fit |
|---|---|
| A developer who wants a cheap box and can wait | Wait for the NVIDIA listing |
| Wants a Spark-class system now and can set it up | A partner GB10 system |
| Not sure yet whether local hardware is worth it | Rent a cloud GPU first |
| A business that needs a working private AI system, not a project | Turnkey, with a written delivery date |
For a deeper checklist before you buy anything, see our local LLM hardware buying checklist and the DGX Spark buyer’s guide.
What we could not verify
- Today’s stock and price on NVIDIA Marketplace: the page did not load; the last readable observation is from October 6.
- Partner availability and 64 GB prices: only NVIDIA’s announced starting price is documented.
- Aradia’s current availability of units and delivery dates: not stated on its pricing page.
- Real-world performance and support quality: not tested.
Sources and verification note
NVIDIA’s product page and Aradia’s pricing page were read on October 23, 2026. NVIDIA Marketplace figures are from October 6, and NVIDIA’s 64 GB announcement from October 2 was read on October 7. Prices, stock and terms change; request a current written quote before you buy.
Affiliate disclosure: AI Compute Scout may earn a commission if you purchase through a marked Aradia referral link. This does not affect the analysis or the recommendation, and we list alternatives to Aradia on this page.