For compute buyers

A dedicated GPU. Billed by the second.

  • A whole GPU, never shared.
  • SSH, batch jobs or inference endpoints.
  • Prepaid credit. No credit line.
New instanceSample
GPU
H100 80GB SXM
Region
Americas
Waiting for request--:--:--
  1. Request
  2. Match
  3. Hold
  4. Place
  5. Run

01Choose

Pick the GPU. HZ finds the host.

  • Filter by model, memory and region.
  • Set a price limit.
  • Placed on a healthy GPU that fits.
Available capacitySample data
GPUMemoryRegionsAvailableFrom / GPU-h
H100 SXM80 GBAmericas · Europe · Asia-Pacific41$1.89
H200 SXM141 GBAmericas · Europe18$2.69
B300288 GBAmericas · Asia-Pacific9$4.20
A100 SXM80 GBAmericas · Europe27$1.29

Prices are set by each operator. Availability changes through the day.

02Sourcing

Need more than what's listed? We'll source it.

  • Larger counts or a specific GPU.
  • A region the catalog doesn't cover.
  • A quote before anything runs.
  1. 01

    Tell us

    The GPU, how many and when.

  2. 02

    We source

    HZ's team finds the capacity.

  3. 03

    You run

    A quote first, then your GPUs go live.

Capacity request
Opens an email to sales@hz-ai.io

03Run

Run it your way.

  • Through the dashboard, the API or SSH.
  • Stop any time. Pay for what ran.
  1. On-demand

    SSH

    Full SSH access with your own public key.

  2. Batch

    Jobs

    Run to completion. Pay for the runtime.

  3. Inference

    Endpoints

    A per-deployment gateway and API keys.

04Regions

Capacity on five continents.

  • North and South America, Europe, Asia and Oceania.
  • One account for every region.
Connected sitesRepresentative

05One dashboard

Every data center. One dashboard.

  • Run on any connected site from one account.
  • One prepaid balance and one bill for every region.
  • Monthly spend cap, low-balance alerts and CSV export.
Buyer console · all sitesSample data
Active GPUs42
GPU-hours · 30d21,780
Spend · 30d$41,809
Prepaid balance$12,480
SiteContinentGPUsGPU-h · 30dShare of spendSpend · 30d
DC1 · OregonNorth America84,120$7,787
DC3 · DallasNorth America42,050$3,875
DC2 · AshburnNorth America126,300$11,907
DC8 · São PauloSouth America2980$1,960
DC4 · FrankfurtEurope63,110$6,066
DC5 · StockholmEurope21,020$1,836
DC6 · SingaporeAsia42,240$4,368
DC7 · TokyoAsia21,060$2,120
DC9 · SydneyOceania2900$1,890
Monthly spend cap · all regions$41,809 of $50,000 · 84%

06Billing

Only pay for what runs.

  • A hold before the job starts.
  • Settled to actual runtime.
  • Unused credit stays in your account.
Open-ended sessionSample

Running · H100 80GB SXM · DC2 Ashburn

Elapsed00:12:34
Billed so far$0.41
Credit balance$1,240.17
Can run for656 h 10 m

$1.89 / GPU-h, billed by the minute from your prepaid credit.

  • Settled windows
  • Current window
  • Held ahead
  1. 01

    Start

    Credit for a short window is held before it runs.

  2. 02

    Keep running

    The next window is held just before the current one ends. There's no end time to set, and it stops if your credit runs out.

  3. 03

    Stop any time

    The current minute settles. Anything held ahead returns to your balance.

07Isolation

Isolated workloads, near bare-metal performance.

  • One dedicated GPU, never shared.
  • Private networks and cloud metadata blocked.
  • Inference through a per-deployment gateway.
One node, sandboxLive
Isolation sandboxBuyer GPUShared host kernel, filteredHZ gateway · mTLSBuyer SSH · directPrivate networksCloud metadataOther buyers✕✕✕

08FAQ

Questions buyers ask.

How is billing calculated?

You prepay credit. Before a job starts, HZ places a hold. When it ends, the hold settles to the actual runtime at the node's GPU-hour rate, with a 60-second minimum. Unused credit stays in your account.

Can I SSH into an instance?

Yes. On-demand instances give full SSH access with your own public key. You can also run batch jobs or deploy inference endpoints.

How isolated is my workload?

Each workload runs on a dedicated GPU in its own isolated environment, with private networks and cloud metadata blocked. Inference traffic goes through a per-deployment gateway.

Which GPUs can I get?

Whatever connected operators have available. Availability tracks real capacity, so it changes through the day.

09Start

Get your first GPU.