GPUBank vs Novita AI
Novita AI is an AI cloud that leads with serverless model APIs and agent sandboxes, and also sells per-second GPU container instances, serverless GPU endpoints and 8-GPU bare metal nodes; its terms name no legal entity beyond Novita AI. Its H100 SXM costs $3.39 per GPU-hour on demand and $1.70 on spot; GPUBank charges $1.85 on demand. Novita has serverless GPUs, SOC 2 Type II and PayPal; GPUBank has a wider on-demand catalogue, crypto top-ups and named regions.
GPUBank
Tallinn, Estonia
H100 SXM
$1.85 / h
- vCPU, RAM, NVMe, IPv4 and traffic included
- Per-second billing after the health check
- Top up in Bitcoin, USDT and 12+ other coins
vCPU, RAM, NVMe, IPv4 and traffic included
Novita AI
GPU cloud
H100 SXM
$3.39 / h
- Serverless GPU endpoints with autoscaling and per-second worker billing
- 8x H100 SXM bare metal at $1.70 per GPU-hour, with NVLink and RDMA
- SOC 2 Type II, report under NDA, and a DPA with SCCs
Novita AI
GPU cloud
24 points compared25 of their pages readprices last checked on 29 September 2026page updated 2 November 2026
The verdict
Serverless GPUs and SOC 2 at Novita AI. Lower on-demand prices and crypto top-ups at GPUBank.
Novita suits teams that already use its model APIs and want GPU containers, serverless endpoints and spot capacity on the same balance. Its public GPU list has five instance models, and for the GPUs both of us rent on demand GPUBank costs up to 45 % less, with the L40S at the same $0.55. Novita's own pages disagree on some prices, for example the RTX 5090 at $0.56 or $0.73 on demand. Pay-as-you-go instances release their hardware when stopped and may not restart, and stopped data is deleted after 7 days.
Novita AIGPUBank 01The real cost
The same GPUs, side by side, per hour.
Every GPU model both of us rent, at each provider's own published price for one card, in US dollars, excluding VAT. Prices last checked on 29 September 2026.
| GPU | Novita AI per hour | GPUBank per hour | Difference | 100 h at Novita AI | 100 h at GPUBank |
|---|---|---|---|---|---|
 B200 SXM180 GB · their reserved price |
$4.77 | $4.15 | −13 % | $477.00 | $415.00 |
 H100 SXM80 GB |
$3.39 | $1.85 | −45 % | $339.00 | $185.00 |
 L40S48 GB |
$0.55 | $0.55 | −0 % | $55.00 | $55.00 |
 RTX 509032 GB |
$0.56 | $0.48 | −14 % | $56.00 | $48.00 |
 RTX 409024 GB |
$0.33 | $0.23 | −30 % | $33.00 | $23.00 |
Novita's B200 figure is a reserved bare metal rate on 8-GPU nodes, sold through sales. Its pages disagree on some prices; the RTX 5090 shows $0.56 or $0.73 on demand.
Where Novita AI publishes no on-demand price for a model, its lowest published reserved price is used, as labelled on the row.
A fine-tune, 40 h on an H100 SXM
Novita AI
$135.60
GPUBank
$74.00
Saves $61.60 at GPUBank (−45 %)
An inference box, a month, 720 h, on an L40S
Novita AI
$396.00
GPUBank
$396.00
Saves $0.00 at GPUBank (−0 %)
Image generation, 200 h on an RTX 4090
Novita AI
$66.00
GPUBank
$46.00
Saves $20.00 at GPUBank (−30 %)
02Their catalogue
Every GPU Novita AI lists.
7 GPU models listed by Novita AI, each at its lowest public price per GPU-hour. 6 of them are also rented at GPUBank (in blue).
- B200192 GB · reserved$4.77
H200141 GB · reservedNo price- H100 SXM80 GB · on demand$3.39
- L40S48 GB · on demand$0.55
RTX 6000 Ada48 GB · on demand$0.77- RTX 509032 GB · on demand$0.56
- RTX 409024 GB · on demand$0.33
03What the price includes
What appears on the bill besides the GPU.
The same H100 SXM session of 2 h 14 m 03 s, line by line, as each provider describes its billing.
Novita AI
2 h 14 m 03 s on H100 SXM
- GPU · $3.39 / h
- $7.57
- vCPU and RAM
- Included
- Storage
- Network volume $0.06/GB-month
- Egress
- Included
- Public IPv4
- Port mapping; dedicated IP on request
- Billing granularity
- Per second, settled hourly
- Minimum charge
- Not published
GPUBank
2 h 14 m 03 s on H100 SXM
- GPU · $1.85 / h
- $4.13
- vCPU and RAM
- Included
- Storage
- Local NVMe included; persistent disk $0.03 per GB-month
- Egress
- Included
- Public IPv4
- Included
- Billing granularity
- Per second, from the health check
- Minimum charge
- None
While stopped
What keeps running on the bill.
Novita AI
Disk kept 7 days at $0.005/GB-day.
Then deleted. Pay-as-you-go hardware is released and may not be available to restart.
GPUBank
Only the persistent disk, $0.03 per GB-month. Compute, IPv4 and scratch cost nothing.
04Where the GPUs run
Where Novita AI runs GPUs is not published.
Countries where Novita AI says its GPUs run, from its own region list or documentation, against GPUBank's four regions. A pin marks a country, not the exact site.
Paris
Frankfurt
Ashburn
Singapore
Novita AI0
No country list on its public pages.
GPUBank4
Paris
Frankfurt
Ashburn
Singapore
Novita AI · 0 countriesGPUBank · 4 regions
05Paying and signing up
How each one takes your money.
Novita AI
Payments and sign-up
Payment methods
Cards
No cryptoMinimum top-up More than $10
To start
- 1Email and password
Billing modelPrepaid balance
GPUBank
Payments and sign-up
Payment methods
A prepaid USD balance, topped up in crypto. Every GPU-second is drawn from it.
Minimum top-up $25
To start
- 1Email and password
- 2Top up in crypto
- 3Launch a GPU
Billing modelPrepaid balance
06Trust and support
Certificates, uptime and who answers.
| Novita AI | GPUBank | |
|---|---|---|
| SOC 2 | Type II | Not published |
| ISO 27001 | Not published | Not published |
| HIPAA | Not published | No |
| GDPR DPA | DPA available | Published DPA, applies automatically |
| Availability target | No figure; contractual SLAs on bare metal | 99.9 % monthly, per region |
| Status page | status.novita.ai | Public status page |
| Support | Email, Discord, sales; 24/7 claimed | Tickets from a $25 balance |
07Their site, our site
What you see before you pay.
Both sites as a visitor sees them, without an account: the Novita AI pricing page captured on 29 September 2026, and GPUBank's live board. Cookie banners and chat bubbles are closed; nothing else is edited.
- 1The pricing page opens on the model API tab; GPU prices sit under a separate GPUs tab, visible without an account.
- 2The console explore page showed RTX 4090, RTX 6000 Ada and RTX 5090 as Unavailable, H100 SXM stock Low and L40S High.
- 3The Spot landing page prices differ from the pricing page for RTX 5090 and L40S.
- 1Live stock per region and every price, without an account.
- 2Per-hour and per-second prices, with vCPU, RAM, NVMe, IPv4 and traffic in them.
- 3A machine-readable price feed at /prices.json.
08The spec sheet
24 points, read off both sites.
GPUBank aheadNovita AI aheadEven
| Criterion | GPUBank | Novita AI |
|---|---|---|
| Price and catalogue | ||
| H100 SXM, per GPU-hourSpot has a 1-hour guaranteed runtime, then 1-hour notice. GPUBank interruptible is $1.11. | $1.85, all-in | $3.39 on demand, $1.70 spot |
| Cheapest GPU-hourRTX 4090 spot is $0.17 in the price table and $0.18 in a Spot page headline. | $0.23, RTX 4090 | $0.33, RTX 4090 |
| GPU models listedFive as instances; B200 and H200 only as bare metal. | 10 | 7 |
| What the GPU price includesNo public IPv4 by default. | vCPU, RAM, NVMe, IPv4, traffic | vCPU, RAM, 60 GB disk, traffic |
| Billing | ||
| Billing granularityBilling starts when the instance enters the Pulling state. | Per second | Per second, settled hourly |
| Minimum charge | None | Not published |
| Stopped instanceThen deleted. Pay-as-you-go hardware is released and may not be available to restart. | Disk only | Disk kept 7 days at $0.005/GB-day |
| Persistent storageBound to one data center. Local volume disks cost $0.005 per GB-day. | $0.03 per GB-month | Network volume $0.06/GB-month |
| Egress traffic | Included | Included |
| Public IPv4Up to 15 exposed ports, tools included. | Included | Port mapping; dedicated IP on request |
| Billing model | Prepaid balance | Prepaid balance |
| Payment and sign-up | ||
| Crypto paymentsThe FAQ states that cryptocurrency payments are not currently supported. | Bitcoin, USDT, Ethereum and 12+ more | Not supported |
| Card paymentsPayPal is processed manually within 7 business days. | No | Cards via Stripe, PayPalAhead |
| Minimum deposit | $25 | More than $10Ahead |
| Platform | ||
| API and CLI | REST API, CLI, Python SDK | REST API; CLI documented for Sandbox only |
| Terraform | Official provider | Claimed, not documented |
| Serverless GPUs | No | Serverless GPU endpointsAhead |
| Managed Kubernetes | No | Not published |
| Ready-made templates | 25 | Templates, count not published |
| Interruptible or spot | Yes, 40 % off | Yes, 1-hour guaranteed runtime |
| GPUs per instance | Up to 8 | Up to 8 per instance |
| Multi-node interconnectNot published for GPU instances. | Not offered | NVLink and RDMA on bare metalAhead |
| Network and regions | ||
| RegionsMarketing claims 14 global regions in one place and 20+ locations in another. | 4 | Not published |
| Countries | France, Germany, US, Singapore | Not published |
| EU regions | Paris, Frankfurt | Not published |
| Trust and support | ||
| Availability targetNo SLA found for GPU instances. | 99.9 % monthly | No figure; contractual SLAs on bare metal |
| Compliance reports | None published | SOC 2 Type IIAhead |
| Public status page | Yes | status.novita.ai, no GPU component |
| SupportNamed TAM on Enterprise plans. | Tickets from a $25 balance | Email, Discord, sales; 24/7 claimedAhead |
09The right fit
Different needs. A clear choice.
What Novita AI does better
- Serverless GPU endpoints with autoscaling and per-second worker billing
- 8x H100 SXM bare metal at $1.70 per GPU-hour, with NVLink and RDMA
- SOC 2 Type II, report under NDA, and a DPA with SCCs
- Pay by card or PayPal; RTX 6000 Ada, which GPUBank does not list
Choose GPUBank if
- H100 SXM at $1.85 on demand, against $3.39 at Novita
- B200, H200, H100 PCIe, A100, RTX PRO 6000 and MI300X on demand
- Stop an instance and keep its disk; Novita deletes it after 7 days
- Pay in Bitcoin, USDT and 12+ other coins; Novita takes no crypto
- Public IPv4 included and four named regions
Moving from Novita AI takes an afternoon.
Novita instances run Docker images, so the same image runs on GPUBank. Network volumes have no documented export tool, and transfer is free, so copy data out over SSH before stopped data is deleted after 7 days.
- 1Copy your volume data from the Novita instance with rsync or scp over SSH.
- 2Launch the same GPU on GPUBank with a persistent disk and run the same Docker image.
- 3Stop the Novita instance and release its network volume.
# 1. a GPU and a disk on GPUBank gpubank launch --gpu h100-sxm --region eu-paris --disk data # 2. copy your data from the old machine rsync -azP old-host:/workspace/ root@<new-ip>:/workspace/ # 3. run the same image or template ssh root@<new-ip> 'cd /workspace && ./start.sh'
A few things you might be wondering.
On demand, yes. Novita lists the H100 SXM at $3.39 per GPU-hour on demand and $1.70 on spot. GPUBank charges $1.85 on demand and $1.11 on interruptible instances, with vCPU, RAM, NVMe, IPv4 and traffic included.
No. Novita's FAQ states that cryptocurrency payments are not currently supported; it takes cards through Stripe and PayPal. GPUBank takes Bitcoin, USDT, Ethereum and more than a dozen other coins, and no cards.
Novita's official pages disagree on some figures. The RTX 5090 is $0.56 on demand on the pricing page and console, and $0.73 on the Spot page; the RTX 4090 spot rate is $0.17 or $0.18. The docs say console prices are for reference only and the actual order prevails.
Yes. Novita sells serverless GPU endpoints with autoscaling and per-second worker billing, with prices shown only in the logged-in console. GPUBank has no serverless product and rents instances by the second.
Compute billing stops, but data is kept only 7 days at $0.005 per GB-day and then deleted, and pay-as-you-go hardware is released and may not be available on restart. On GPUBank a stopped instance keeps its persistent disk at $0.03 per GB-month.
Without. Every price on this page excludes VAT and sales tax, for Novita AI and for GPUBank, so the two columns compare like for like. Tax is added at checkout according to where the buyer is based.
Sources
Every figure about Novita AI was read on its own pages or official documentation. Prices last checked on 29 September 2026.
- 01novita.ai/pricingGPU instance on-demand and spot prices, 1x and 8x, storage prices
- 02novita.ai/gpusGPU price cross-check, platform claims (regions, Terraform, CLI, support, serverless)
- 03novita.ai/gpus-console/explorePer-card vCPU/RAM, stock, Secure Cloud badges, on-demand prices
- 04novita.ai/gpus-spotSpot prices, spot policy (1-hour protection, 1-hour notice)
- 05novita.ai/gpu-baremetalBare metal H100/B200 prices, H200/RTX configs, interconnect
- 06docs.novita.ai/guides/gpu-instance-pricingBilling formula, per-second billing, storage prices
- 07docs.novita.ai/guides/gpu-instance-faqStopped instance retention, multi-GPU, subscriptions
- 08docs.novita.ai/guides/gpu-instance-overviewFree network transfer, templates, per-second billing
- 09docs.novita.ai/guides/gpu-instance-quickstart-manage-network-volumeNetwork volume billing and data center binding
- 10docs.novita.ai/guides/serverless-gpus-pricingServerless GPU billing method
- 11docs.novita.ai/guides/serverless-gpus-overviewServerless features
- 12docs.novita.ai/api-reference/gpu-instance-create-instance.mdBilling modes (onDemand, monthly, spot), max 8 GPUs, VPC, network storage
- 13docs.novita.ai/api-reference/gpu-instance-list-products.mdProduct fields incl. monthlyPrice, spotPrice, regions
- 14docs.novita.ai/guides/payment-methodsStripe, minimum top-up, auto top-up
- 15docs.novita.ai/guides/faqPayment methods (card, PayPal, no crypto), refunds, dedicated IP, restart behaviour
- 16docs.novita.ai/guides/quickstartSign-up methods, verification mention, Discord support
- 17novita.ai/user/registerSign-up options, Sandbox credits
- 18docs.novita.ai/guides/gpu-instance-quickstart-connect-to-instanceSSH access
What we could not verify (8)
- Monthly/yearly subscription prices: exist per API docs (monthlyPrice, 1 to 12 months) but not shown publicly
- Serverless GPU per-second worker prices: only visible in the logged-in Serverless console
- Legal entity name, headquarters, founding year and funding: not found on official pages read
- Terraform: claimed on the /gpus page ("REST, gRPC, Terraform and a powerful CLI") but no Terraform provider documented; CLI documentation covers Agent Sandbox only
- A100 appears in a "GPU Types Available" list on the Spot page with no price or console listing; not counted in catalogue_count
- Wire transfer/invoicing for self-serve accounts not published
- Whether cards without the Secure Cloud badge run on third-party hosts is not documented
- Bare metal CPU/RAM/storage per node and commitment term not published
In short
GPUBank vs Novita AI, in facts.
The essentials of this comparison in one table, for quick readers and AI assistants. Prices last checked on 29 September 2026.
Read as Markdown- H100 SXM, per GPU-hour
- Novita AI: $3.39 on demand, $1.70 spot
GPUBank: $1.85, all-in - Cheapest GPU-hour
- Novita AI: $0.33, RTX 4090
GPUBank: $0.23, RTX 4090 - Billing granularity
- Novita AI: Per second, settled hourly
GPUBank: Per second - Stopped instance
- Novita AI: Disk kept 7 days at $0.005/GB-day
GPUBank: Disk only - Persistent storage
- Novita AI: Network volume $0.06/GB-month
GPUBank: $0.03 per GB-month - Egress traffic
- Novita AI: Included
GPUBank: Included - Crypto payments
- Novita AI: Not supported
GPUBank: Bitcoin, USDT, Ethereum and 12+ more - Card payments
- Novita AI: Cards via Stripe, PayPal
GPUBank: No - Minimum deposit
- Novita AI: More than $10
GPUBank: $25 - GPU models listed
- Novita AI: 7
GPUBank: 10 - Regions
- Novita AI: Not published
GPUBank: 4 - Serverless GPUs
- Novita AI: Serverless GPU endpoints
GPUBank: No - Availability target
- Novita AI: No figure; contractual SLAs on bare metal
GPUBank: 99.9 % monthly - Compliance reports
- Novita AI: SOC 2 Type II
GPUBank: None published - Support
- Novita AI: Email, Discord, sales; 24/7 claimed
GPUBank: Tickets from a $25 balance - Headquarters
- Novita AI: Not published
GPUBank: Tallinn, Estonia
How this page was made
Every price and policy of Novita AI was read on its own website, documentation or public API on 29 September 2026, with the address of each page listed above. A value we could not read on a primary source is marked "Not published" rather than estimated. GPUBank figures come from this site and its terms. Prices change: check both providers before you buy. Novita AI, its logo and its website belong to their owner and are shown only to identify the service compared. GPUBank is not affiliated with Novita AI.
