DEDICATED AI COMPUTE. SIMPLIFIED.

Rent a whole DGX Spark,
billed by the minute.

Your own NVIDIA DGX Spark. No shared GPU, no complicated setup. Build, test, and run AI workloads on a dedicated machine—on your terms.

From $0.85 / hour
Full-machine access  ·  Prepaid credits  ·  Transparent billing
YOUR DEDICATED MACHINE
Illustrative studio render of NVIDIA DGX Spark dedicated AI computer
GB10Blackwell
✦128 GBUnified memory
MACHINENVIDIA DGX Spark
100% dedicated
MOVE CURSOR TO EXPLORE
128 GBUnified memory
20 coresArm CPU
1 PFLOPUp to FP4 AI performance*
1 deviceEntirely yours

Powerful compute.
Predictable pricing.

One dedicated DGX Spark. A straightforward hourly rate, metered by the minute. Add credits to your account and manage your machine from the VLAN24 Spark portal.

Whole-machine rental Minute-based metering Prepaid account credits
DGX Spark · DedicatedSTARTING RATE
₩1,130/ hour

Based on the portal's lowest price$0.85 USD / hr

KRW is an estimate at ₩1,330 / $1 (19 Nov 2026), not a live currency quote. Final price and currency are confirmed in the user portal.

4 hours
1 hour48 hours

South Korea's standard VAT is 10%, but whether it applies depends on the supplier and transaction. Tax shown here is only an illustration, not tax advice.

Estimated rental cost₩4,522≈ $3.40 USD
At the reference FX rate
Start with VLAN24 Spark
Important: Billing continues even when your machine is stopped. Charges apply for the full rental period.

Small footprint.
Serious AI power.

Build on NVIDIA Grace Blackwell architecture with a dedicated system designed for AI development, inference, and fine-tuning.

PLATFORMNVIDIA GB10

Grace Blackwell Superchip with integrated Blackwell GPU.

GB10
MEMORY128 GB

LPDDR5x coherent unified system memory.

PROCESSOR20-core Arm

10 Cortex-X925 + 10 Cortex-A725 CPU cores.

STORAGEUp to 4 TB

NVMe M.2 SSD; capacity depends on configuration.

CONNECTIVITY10 GbE

ConnectX-7 networking, Wi-Fi 7 and USB-C.

AI PERFORMANCEUp to 1 PFLOP

Theoretical FP4 throughput with sparsity.

* Specifications are based on NVIDIA documentation. Storage and service configuration must be confirmed before launch. Peak FP4 performance is theoretical.

Your next AI idea
starts here.

One dedicated system for exploring models, testing workflows, and building useful applications.

01 / AI

Prototype AI models

Experiment, validate ideas, and iterate on AI projects with your own machine.

02 / INFERENCE

Run inference

Test responses and workflows on dedicated hardware without sharing compute.

03 / DEVELOPMENT

Build and test

Develop and debug AI workflows through your remote environment.

Ready to turn your idea into a working prototype?Create an account to begin.
Get started

From signup to
supercomputing.

Three simple steps to start using your own DGX Spark.

01

Create an account

Sign up with your email in the existing VLAN24 Spark user portal.

Sign up
02

Add prepaid credit

Sign in, open Billing, and top up the balance that pays for your rental.

Open user portal
03

Request your machine

Go to Deploy in the portal to request a dedicated DGX Spark.

Sign in to deploy

Quick answers.

A few details worth knowing before you rent.

Will I be billed while my machine is stopped?

Yes. Billing continues for the entire rental period, including when the machine is stopped. Stopping a machine does not pause charges.

Is this an entire machine or a shared GPU?

This service is designed for dedicated rental of a whole DGX Spark, not a shared fraction of a GPU.

Is VAT automatically added to the price?

Not necessarily. Korea generally has a 10% VAT rate, but actual applicability depends on the transaction. You can add an illustrative 10% in the calculator.

Have questions?
Let's talk Spark.

Need help with pricing, availability, or getting started? Reach out to our team.

Email our teamhelp@vlan24.com