Zettabyte review
Updated September 28, 2026 · DeployCue Editorial Team · Facts checked September 26, 2026
We earn a commission when you sign up through links on this site. Partner status can affect where a brand appears in our rankings. Partners with a live offer in your country come first, then every other brand by score.
Our verdict
Zettabyte suits teams that need datacenter NVIDIA parts, mainly H100, H200, B200 and GB200, and want one hourly on-demand rate. The catch is cost: the GPU hour from $2.20 lands well above the €0.63 median, so short jobs pay for idle headroom.
Key facts
- Best for
- Single-model NVIDIA GPU rental with hourly on-demand billing.
- GPU models
- NVIDIA H100, NVIDIA H200, NVIDIA B200, NVIDIA GB200↗
- GPU hour from
- $2.20↗
- Billing granularity
- Hourly↗
- Legal entity
- Zettabyte Holdings Inc.↗
- Headquarters
- Taiwan↗
- Support channels
- Email, Phone, Help-centre↗
- Pricing model
- Hourly billing
- Official website
- zettabyte.com
Pros and cons
Pros
- Top datacenter parts on one menu, including NVIDIA H100, H200, B200 and GB200 in NVIDIA H100, NVIDIA H200 and NVIDIA B200
- Entry price of $2.20 for an on-demand GPU compute instance on hourly billing
- Flexible commitment through hourly and reserved, covering hourly and reserved options
- Uptime commitment of 99.9% for running jobs
Cons
- The GPU hour from $2.20 runs above the €0.63 median, so long jobs get expensive
- Billing granularity is hourly, and a short job still pays a full hour
- Support runs on email, phone and help centre only, with no live chat listed
Who it suits
- Teams training or serving on datacenter NVIDIA silicon rather than consumer cards
- Developers who want to start small on an on-demand instance at $2.20 before committing to reserved capacity
- Anyone who values 99.9% uptime for long-running jobs
Who should look elsewhere
- Budget-first teams: ComputePrices.com lists a wider set in H100 SXM, A100 SXM and H200, including consumer cards, so more price points to weigh
- Builders who need AMD or non-NVIDIA options, since Salad covers selected AMD classes in NVIDIA RTX 3060, NVIDIA RTX 3070 and NVIDIA RTX 3070 Ti alongside RTX cards
- Short burst workloads that need per-second billing, since Zettabyte bills hourly
The company behind it
Zettabyte is run by Zettabyte Holdings Inc., registered in Taiwan, so contracts and support hours are anchored outside the US. Support runs on email, phone and help centre, covering email, phone and a help centre, with no live chat in the list. The service is GPU focused, with hourly billing as the only listed model.
How Zettabyte compares
Zettabyte vs RunPod8.7 vs 9.5
Zettabyte vs ScalaHosting8.7 vs 8.3
Zettabyte vs Contabo8.7 vs 9.2
Zettabyte vs Kamatera8.7 vs 9.5
Zettabyte vs Cloudways8.7 vs 9.2
Zettabyte vs PlanetHoster8.7 vs 8.7
Zettabyte vs Acronis8.7 vs 9.9
Zettabyte vs Aruba8.7 vs 9.3
Zettabyte vs Hostinger8.7 vs 9.1
Zettabyte vs Decodo8.7 vs 9.1
Zettabyte vs MarsProxies8.7 vs 9.0
Zettabyte vs hide.me8.7 vs 8.7
Zettabyte vs Proton Drive8.7 vs 8.3
Frequently asked questions
What does an hour of GPU compute cost?+
GPU compute starts at $2.20 for the cheapest listed card. That is well above the €0.63 median across providers on the same measure, so treat the floor rate as a premium on typical market pricing.
How are GPU jobs billed?+
Billing granularity is hourly, and hourly lists hourly as the only pricing model. Available billing models cover hourly and reserved, so a reserved arrangement is possible for steady workloads, while short jobs stop at the hourly mark.
Which GPUs can I rent?+
The listed range in NVIDIA H100, NVIDIA H200 and NVIDIA B200 covers NVIDIA H100, H200, B200 and GB200, all datacenter parts. There is no consumer RTX or AMD line in that list, which narrows choice for lighter or mixed fleets.
What does it cost to get started?+
The entry point is $2.20, described as an on-demand GPU compute instance with hourly billing. That sits under the €5.83 median across the comparison, so a first run is approachable even though the hourly floor runs high.
What uptime is committed?+
The stated uptime commitment is 99.9%. For long training runs that number is the baseline to hold the provider to, and it falls short of the highest tier many buyers expect from hyperscalers.
Who is Zettabyte suited for?+
The lineup in NVIDIA H100, NVIDIA H200 and NVIDIA B200 points to teams already committed to NVIDIA datacenter accelerators for training or inference. It suits buyers who want a short on-demand trial at $2.20 before moving to reserved capacity under hourly and reserved.
How do I reach support?+
Support runs on email, phone and help centre, which covers email, phone and a help centre. Chat is not among them, so live, in-conversation help is not part of the listed support model.
Can I lock in capacity for a longer job?+
Billing models cover hourly and reserved, so a reserved arrangement sits alongside the hourly on-demand model of hourly. The on-demand entry point is On-demand GPU compute instance, hourly billing at $2.20, so a short run can start before a longer commitment.
Sources
- zettabyte.com/terms checked September 26, 2026
Backs: Billing granularity, Billing models, Entry plan, Headquarters, Legal entity, Support channels
“Usage charges for On-Demand instances are calculated by the hour (or fraction thereof).”
- zettabyte.com/zcloud/pricing checked September 26, 2026
Backs: Entry price
“Reserved from $2.20 / GPU-hour”
- zettabytecloud.com checked September 26, 2026
Backs: SLA uptime
“99.9% uptime SLA target”
- zettabytecloud.com/pricing checked September 26, 2026
Backs: GPU hour from, GPU models, Pricing model
“On-Demand H100 from $2.51 / Reserved H100 from $2.20 / Private Cloud”
Compiled from 4 source pages, checked September 26, 2026. We did not open an account.
Not confirmed yet: Egress fee.