Rent a dedicated DGX Spark.

One whole NVIDIA DGX Spark in EU-Central, billed per minute from prepaid credit.

Create your account

Connecting to GPUwerk…

Checking capacity…

GPUwerk console demo

This is the offline demo of console.gpuwerk.com. Everything below runs from this one file, state lives in your browser localStorage, and nothing here reaches a real server.

See pricing

EU-Central · checking availability · first top-up: pay $20, get $30 in credit

One whole machinenever shared with another renter
Key-based SSH accessno password login for SSH
EU data residencyhardware in EU-Central
Status published openlystatus.gpuwerk.com

The hardware

Every instance is one whole machine. No other renter shares it.

NVIDIA DGX Spark · per node
Superchip
GB10 Grace Blackwell
AI performance
1 PFLOPS FP4 (sparse)
Memory
128 GB unified LPDDR5x
CPU
20-core Arm
Storage
1 TB NVMe
Network
ConnectX-7 · 200 Gb/s
OS
DGX OS 7 · Ubuntu 24.04
Preparation
Readiness checked before billing

What fits in 128 GB

Unified memory is the whole point, a single node holds models that would otherwise need several discrete GPUs.

Pricing

Per-minute billing, $0 setup fee, no egress charges. Stopping holds your Spark at a discount off the hourly rate; terminating is what stops billing. First top-up: pay $20, get $30 in credit. Refer a friend: when their first top-up earns the bonus, you get the same amount, up to $10. One per payment method: a top-up paid with your card, or with a card another referral already used, doesn't qualify.

Recommended

On-Demand

$0.79/hour
Nodes
1× DGX Spark
Memory
128 GB
Storage
1 TB NVMe
Setup fee
$0
Commitment
None

A month, always on

$576.70/month
What it is
$0.79/h × 730 hours
Nodes
1× DGX Spark
Up front
Nothing
Setup fee
$0
Commitment
Stop any time

Docs: deploy and connect

Every instance is a fresh container on its own machine, with a hostname and key-based SSH. You are root inside your container, not on the host; bring your own stack. Managed model serving is coming soon.

# deploy from the API
curl -X POST https://api.gpuwerk.com/v1/instances \
  -H "Authorization: Bearer $GPUWERK_TOKEN" \
  -d '{"image":"dgx-os","ssh_public_key":"ssh-ed25519 AAAA..."}'

# your shell in your container
ssh [email protected]

# the whole GB10 is yours
nvidia-smi && python3 --version

# locked yourself out? reinstall every key on the account, no rebuild needed
curl -X POST https://api.gpuwerk.com/v1/instances/<id>/reset-access \
  -H "Authorization: Bearer $GPUWERK_TOKEN"

FAQ

The short version of how billing and instances work.

How does billing work?

Per-minute, against a prepaid credit balance. Stopping an instance holds its Spark for you at a discount off the hourly rate; terminating releases the hardware and is what actually stops billing. Your first top-up of at least $20 earns a $10 bonus (pay $20, get $30 in credit). If your balance runs out, your instances are stopped and their Sparks released. Each workspace stays on its Spark for up to 48 hours: top up and start it again in that time and it picks up where it left off, unless someone else has rented that Spark, which deletes it sooner. Workspaces are never backed up, so keep your own copy. Nothing is silently overdrafted.

What happens to my data when I stop vs. terminate?

Stopping keeps everything on the node's NVMe, start it again and your files, models, and environment are exactly as you left them. Terminating releases the hardware and permanently erases local data, which is why it requires typing the instance name to confirm.

Is there an egress or ingress fee?

No. The hourly rate is the whole price, bandwidth in and out over the 200 Gb/s ConnectX-7 link is included.

How isolated is my instance?

Each instance is one whole physical DGX Spark, no virtualization, no shared GPUs, no noisy neighbors. SSH access is key-based only; password login for SSH is disabled across the fleet. Optional VNC desktop access, if you enable it at deploy time, uses a separately generated password for that connection only. Hardware is located in EU-Central.

Do you offer managed model serving?

Not yet, today every instance is a clean box you drive over SSH: install vLLM, SGLang, or whatever stack you prefer, and it all fits in unified memory. Preloaded model images that serve on boot are on the roadmap.

Start building on a Spark today.

First top-up: pay $20, get $30 in credit. Refer a friend: when their first top-up earns the bonus, you get the same amount, up to $10. One per payment method: a top-up paid with your card, or with a card another referral already used, doesn't qualify. Model preparation can take several minutes. $0 setup fee, no commitments, no egress charges.