Operator · San Francisco, CA · founded 2023
Novita AI
AI cloud selling serverless model APIs and agent sandboxes, plus per-second GPU container instances (on-demand, spot, monthly subscription), serverless GPU endpoints and 8-GPU bare metal nodes.
GPU types6with a current offer
Current offers107107 readings on file · last 45 days
H100 from$3.39per GPU-hour, on-demand
Entry price$0.55cheapest GPU-hour listed
Countries1regions not published
Trust Score v0.182/100
- Company & track record47
- Infrastructure control100
- Pricing & billing transparency100
- Reliability signals85
- Security & compliance73
- Customer signal—
Scored Sep 29, 2026 from the provider's own pages. Dimensions not yet measured are excluded from the denominator. Criteria ↓
Outbound link carries a commission. It does not affect this page.
Trust Score · 5 of 6 dimensions measuredThe score, criterion by criterion
Each point comes from something the provider publishes, with the page it was read on and the date. Anything not disclosed earns zero.
Company & track record
weight 15%
47 / 100
| Founding year published | 2023novita.ai · Sep 29, 2026 | 4 / 4 |
| Headquarters published | 🇺🇸 San Francisco, CAnovita.ai · Sep 29, 2026 | 3 / 3 |
| Legal entity named | not disclosednovita.ai · Sep 29, 2026 | 0 / 4 |
| Funding or ownership disclosed | Not published on the pages read. Terms of Service name the contracting party only as "Novita AI" and are governed by Delaware law.novita.ai · Sep 29, 2026 | 0 / 4 |
Infrastructure control
weight 15%
100 / 100
| Capacity model published | operates or rents dedicated capacitynovita.ai · Sep 29, 2026 | 10 / 10 |
| Regions published | 1 regions or countriesnovita.ai · Sep 29, 2026 | 5 / 5 |
Pricing & billing transparency
weight 20%
100 / 100
| Public prices without an account | yes (read on the pricing page)novita.ai · Sep 29, 2026 | 5 / 5 |
| Billing granularity published | per secondnovita.ai · Sep 29, 2026 | 4 / 4 |
| Egress policy published | included (GPU Instance overview: "provides free network transfer")novita.ai · Sep 29, 2026 | 4 / 4 |
| Stopped-instance billing published | Compute billing stops when the instance is stopped. Data of stopped instances is retained 7 days with storage fees of $0.005/GB/day, then deleted permanently. Pay-as-you-go instances release their hardware when stopped and may not restart if it has been reallocated.novita.ai · Sep 29, 2026 | 4 / 4 |
| Public IPv4 policy published | shared/NAT: port mapping on exposed ports (max 15 incl. tools); dedicated IP not open to the public, on requestnovita.ai · Sep 29, 2026 | 3 / 3 |
Reliability signals
weight 20%
85 / 100
| Status page | https://status.novita.ai/novita.ai · Sep 29, 2026 | 6 / 6 |
| SLA published | Bare metal page mentions "contractual SLAs"; no figure published. No SLA for GPU instances found.novita.ai · Sep 29, 2026 | 6 / 6 |
| SLA with a numeric uptime commitment | not disclosednovita.ai · Sep 29, 2026 | 0 / 3 |
| Uptime history published | Status page shows 90-day uptime per component (API Gateway, website, Sandbox, LLM models); no GPU instance component listed. Home page claims 99.5% uptime for model APIs.novita.ai · Sep 29, 2026 | 5 / 5 |
| Observed availability (30 days) | not measured yet | — |
Security & compliance
weight 15%
73 / 100
| SOC 2 | Type IInovita.ai · Sep 29, 2026 | 8 / 8 |
| ISO 27001 | not disclosednovita.ai · Sep 29, 2026 | 0 / 4 |
| HIPAA or GDPR DPA | DPA for enterprise customers, with SCCs (Module 2) and UK Addendumnovita.ai · Sep 29, 2026 | 3 / 3 |
Customer signal
weight 15%
not measured
| Aggregated public reviews | not collected yet | — |
provisional v0.1: evidence from the provider's own pages; dimensions not yet measured are excluded from the denominatorFull method →
Prices by GPU · 5 GPU types with a current offerWhat Novita AI charges, against the market
Cheapest current on-demand true cost here, next to the market median for the same GPU. Negative means cheaper than the market.
| GPU | Cheapest here | Market median | vs market | Current offers | Modes | |
|---|---|---|---|---|---|---|
| H100 | $3.39 | $3.17 | +7% | 10 | on-demandspotreserved | H100 prices |
| L40S | $0.55 | $1.26 | −56% | 3 | on-demandspot | L40S prices |
| RTX 5090 | $0.56 | $0.69 | −19% | 17 | on-demandspotreserved | RTX 5090 prices |
| RTX 4090 | $0.33 | $0.44 | −25% | 11 | on-demandspotreserved | RTX 4090 prices |
| RTX 6000 Ada | $0.77 | $0.79 | −2% | 5 | on-demandspotreserved | RTX 6000 Ada prices |
All offers · 107 readings on fileEvery Novita AI offer, with its receipt
Current offers (last 45 days) ranked by true cost, all rental modes, each with its read date. Older readings stay listed below, in grey.
| GPU | Configuration | Mode | Region | Stock | List | True cost | Read | |
|---|---|---|---|---|---|---|---|---|
| RTX 4090 | 1× RTX 409016 vCPU · 62 GB RAM | Spot | UAE +10 | Unavailable | $0.17 | $0.17 | 21:10 UTCfresh | Receipt |
| RTX 4090 | 1× RTX 409016 vCPU · 96 GB RAM | Spot | US | Unavailable | $0.17 | $0.17 | 21:11 UTCfresh | Receipt |
| RTX 409012-month commitment | 1× RTX 409016 vCPU · 62 GB RAM | Reserved | UAE +10 | Unavailable | $0.27 | $0.27 | 21:10 UTCfresh | Receipt |
| L40S | 1× L40S28 vCPU · 125 GB RAM | Spot | HK | Unavailable | $0.28 | $0.28 | 21:10 UTCfresh | Receipt |
| RTX 40906-month commitment | 1× RTX 409016 vCPU · 62 GB RAM | Reserved | UAE +10 | Unavailable | $0.29 | $0.29 | 21:10 UTCfresh | Receipt |
| RTX 40901-month commitment | 1× RTX 409016 vCPU · 62 GB RAM | Reserved | UAE +10 | Unavailable | $0.30 | $0.30 | 21:10 UTCfresh | Receipt |
| RTX 40901-month commitment | 1× RTX 409016 vCPU · 96 GB RAM | Reserved | US | Unavailable | $0.30 | $0.30 | 21:11 UTCfresh | Receipt |
| RTX 4090 | 1× RTX 409016 vCPU · 62 GB RAM | On-demand | UAE +10 | Unavailable | $0.33 | $0.33 | 21:10 UTCfresh | Receipt |
| RTX 4090 | 1× RTX 409016 vCPU · 96 GB RAM | On-demand | US | Unavailable | $0.33 | $0.33 | 21:11 UTCfresh | Receipt |
| RTX 4090 | 1× RTX 409016 vCPU · 125 GB RAM | Spot | US | Unavailable | $0.34 | $0.34 | 21:10 UTCfresh | Receipt |
| RTX 5090 | 1× RTX 509016 vCPU · 96 GB RAM | Spot | UAE +8 | Unavailable | $0.35 | $0.35 | 21:10 UTCfresh | Receipt |
| RTX 5090 | 1× RTX 509016 vCPU · 120 GB RAM | Spot | UK | Unavailable | $0.35 | $0.35 | 21:10 UTCfresh | Receipt |
| RTX 5090 | 1× RTX 509016 vCPU · 96 GB RAM | Spot | US | Unavailable | $0.37 | $0.37 | 21:10 UTCfresh | Receipt |
| RTX 6000 Ada | 1× RTX 6000 Ada22 vCPU · 125 GB RAM | Spot | US | Unavailable | $0.39 | $0.39 | 21:10 UTCfresh | Receipt |
| RTX 509012-month commitment | 1× RTX 509016 vCPU · 58 GB RAM | Reserved | US | Unavailable | $0.45 | $0.45 | 21:10 UTCfresh | Receipt |
| RTX 50906-month commitment | 1× RTX 509016 vCPU · 58 GB RAM | Reserved | US | Unavailable | $0.48 | $0.48 | 21:10 UTCfresh | Receipt |
| RTX 50901-month commitment | 1× RTX 509016 vCPU · 58 GB RAM | Reserved | US | Unavailable | $0.51 | $0.51 | 21:10 UTCfresh | Receipt |
| L40S | 1× L40S22 vCPU · 125 GB RAM | On-demand | UAE +9 | In stock | $0.55 | $0.55 | 21:10 UTCfresh | Receipt |
| L40S | 1× L40S28 vCPU · 125 GB RAM | On-demand | HK | Unavailable | $0.55 | $0.55 | 21:10 UTCfresh | Receipt |
| RTX 5090 | 1× RTX 509016 vCPU · 58 GB RAM | On-demand | US | Unavailable | $0.56 | $0.56 | 21:10 UTCfresh | Receipt |
| RTX 509012-month commitment | 1× RTX 509016 vCPU · 96 GB RAM | Reserved | UAE +8 | Unavailable | $0.56 | $0.56 | 21:10 UTCfresh | Receipt |
| RTX 509012-month commitment | 1× RTX 509016 vCPU · 120 GB RAM | Reserved | UK | Unavailable | $0.56 | $0.56 | 21:10 UTCfresh | Receipt |
| RTX 40901-month commitment | 1× RTX 409016 vCPU · 125 GB RAM | Reserved | US | Unavailable | $0.59 | $0.59 | 21:10 UTCfresh | Receipt |
| RTX 50906-month commitment | 1× RTX 509016 vCPU · 96 GB RAM | Reserved | UAE +8 | Unavailable | $0.60 | $0.60 | 21:10 UTCfresh | Receipt |
| RTX 50906-month commitment | 1× RTX 509016 vCPU · 120 GB RAM | Reserved | UK | Unavailable | $0.60 | $0.60 | 21:10 UTCfresh | Receipt |
46 fresh offers · the 25 cheapest shown here, all of them in the comparison tool · prices per GPU-hour · true cost includes the add-ons the provider publishesOpen in the comparison tool →
True cost & billing · read Sep 29, 2026What the hourly price does not say
Billing granularity, minimums, what a stopped machine costs, traffic, addresses, deposits and identity checks, as published by the provider.
Billing
- Billing granularity
- per second
- Minimum charge
- not published
- Stopped instance
- Compute billing stops when the instance is stopped. Data of stopped instances is retained 7 days with storage fees of $0.005/GB/day, then deleted permanently. Pay-as-you-go instances release their hardware when stopped and may not restart if it has been reallocated.
- Egress
- included (GPU Instance overview: "provides free network transfer")
- Public IPv4
- shared/NAT: port mapping on exposed ports (max 15 incl. tools); dedicated IP not open to the public, on request
- Billing model
- prepaid credits
- Minimum deposit
- Manual top-up must be greater than $10
- Free credits
- Sign-up page: $100 in Sandbox credits valid 90 days (Agent Sandbox; not stated to apply to GPUs). FAQ mentions a $1 credit voucher revoked on suspicious sign-ups.
Account
- Payment
- card · PayPal
- Sign-up
- email, password
- Identity check
- not published
- Support
- email, discord, sales call
- Enterprise support
- not published
- SLA
- Bare metal page mentions "contractual SLAs"; no figure published. No SLA for GPU instances found.
- Status page
- status.novita.ai ↗
- Published uptime
- Status page shows 90-day uptime per component (API Gateway, website, Sandbox, LLM models); no GPU instance component listed. Home page claims 99.5% uptime for model APIs.
Infrastructure
- Regions
- —
- Countries
- —
- Platform
- API, CLI, serverless, Jupyter, SSH, custom Docker images, persistent volumes, snapshots, spot or interruptible, bare metalSDK: Sandbox SDKs and CLI documented; Serverless GPUs SDK usage guide; OpenAI-compatible LLM APIInterconnect: Bare metal: NVLink 900 GB/s + RDMA (H100/H200 SXM), NVLink 5th gen 1.8 TB/s + RDMA (B200); GPU instances not publishedClusters: 8-GPU bare metal nodes; H100/H200 described with "1000+ GPU linear scaling"; enterprise clusters via sales
- Storage
- ephemeral · Container disk: 60 GB free, excess 0.005 USD/GB-day (0.15 USD/GB-month at 30 days) · Root filesystem; saved with Save Image.volume · Volume disk: 0.005 USD/GB-day (0.15 USD/GB-month at 30 days) · Local, mounted at /workspace; not saved with images.network · Network volume: 0.002 USD/GB-day (0.06 USD/GB-month at 30 days) · Independent lifecycle, bound to one data center, billed per second; released 3 days after the account is in arrears with no running instance.
- Migration out
- Custom Docker images from public or private registries; Save Image pushes the container disk to an image (volume disk excluded). Network volumes persist independently but are bound to one data center and no export tool is documented; data must be copied out over SSH/network (transfer stated as free). Stopped instance data is deleted after 7 days.
Security & compliance
- SOC 2
- Type II
- ISO 27001
- not published
- HIPAA
- not published
- GDPR / DPA
- DPA for enterprise customers, with SCCs (Module 2) and UK Addendum
- Other
- CCPA aligned; SOC 2 Type II report available under NDA to enterprise customers; zero data retention default for API content
- Legal entity
- not published
- Funding
- Not published on the pages read. Terms of Service name the contracting party only as "Novita AI" and are governed by Delaware law.
Read on the provider's own pages · 25 pages read · facts marked "not published" were looked for and not foundPages read ↓
Assessment · documents read Sep 29, 2026Novita AI, in short
What stood out when we read the documents: strengths, weaknesses, who it suits, and what we could not find.
Strengths
- Spot instances at about half the on-demand rate with a 1-hour guaranteed runtime and 1-hour termination notice (H100 SXM $1.70 vs $3.39 per GPU-hour)
- Per-second billing on GPU instances and serverless GPU workers
- Container disk 60 GB free and network transfer included
- Bare metal 8x H100 SXM at $1.70/GPU/hr and 8x B200 at $4.77/GPU/hr published
- SOC 2 Type II audited; DPA with SCCs available
- One account and balance covers model APIs, agent sandboxes, GPU instances and serverless GPU endpoints
Weaknesses
- Only five GPU models on the public GPU instance price list; no A100, H200 or B200 on-demand instances
- Inconsistent prices across official pages (RTX 5090 on-demand $0.56 vs $0.73; L40S spot on one page, absent on another; RTX 4090 spot $0.17 vs $0.18)
- H100 SXM on-demand at $3.39 per GPU-hour
- Monthly subscription prices not public; no region list published
- Pay-as-you-go instances may not restart after stop because hardware is released; stopped instance data deleted after 7 days
- No crypto payment; refunds generally not available
- Dedicated public IP not available by default
Ideal for
- Teams already using Novita model APIs who want GPU containers on the same balance
- Interruption-tolerant jobs on RTX 4090, RTX 6000 Ada or H100 spot capacity
- Serverless inference endpoints billed per second of worker time
- Buyers of 8x H100 or B200 bare metal who can contract through sales
Not published
- Region and data center names/countries: not published (console region filter did not open without login
- API returns cluster IDs only with an API key)
- 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
- Nature of the "required verification" mentioned in the docs quickstart is not described
- no KYC policy published
- 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
Observations
Pricing page opens on the model API tab; GPU prices are under a separate GPUs tab, visible without an account
GPUs tab lists five configurations (H100 SXM, L40S, RTX 4090, RTX 6000 Ada, RTX 5090) with 1x and 8x prices, on-demand and (for three) spot
Console explore page is browsable without login and shows vCPU/RAM per card and stock labels; at check time RTX 4090, RTX 6000 Ada and RTX 5090 showed "Unavailable", H100 SXM "Low", L40S "High"
Only L40S and RTX 5090 cards carry a "Secure Cloud" badge; a "Show Secure Cloud Only" filter and a "Supply GPUs" footer link suggest third-party hosts, not described in docs
GPU page claims "14 global regions" and elsewhere "20+ locations, 4+ continents"; no region names published
Spot landing page prices differ from the pricing page for RTX 5090 and L40S
Bare metal page shows prices for H100 SXM and B200 only; H200, RTX 5090 and RTX 4090 are "Contact us"
Sign-up offers Google, GitHub, Hugging Face or email; headline offer is $100 Sandbox credits
Docs state console prices "are for reference only" and the actual order prevails
Cookie consent dialog blocks the pricing tabs until dismissed
Pages read
- novita.ai/pricing — GPU instance on-demand and spot prices, 1x and 8x, storage prices
- novita.ai/gpus — GPU price cross-check, platform claims (regions, Terraform, CLI, support, serverless
- novita.ai/gpus-console/explore — Per-card vCPU/RAM, stock, Secure Cloud badges, on-demand prices
- novita.ai/gpus-spot — Spot prices, spot policy (1-hour protection, 1-hour notice
- novita.ai/gpu-baremetal — Bare metal H100/B200 prices, H200/RTX configs, interconnect
- docs.novita.ai/guides/gpu-instance-pricing — Billing formula, per-second billing, storage prices
- docs.novita.ai/guides/gpu-instance-faq — Stopped instance retention, multi-GPU, subscriptions
- docs.novita.ai/guides/gpu-instance-overview — Free network transfer, templates, per-second billing
- docs.novita.ai/guides/gpu-instance-quickstart-manage-network-volume — Network volume billing and data center binding
- docs.novita.ai/guides/serverless-gpus-pricing — Serverless GPU billing method
- docs.novita.ai/guides/serverless-gpus-overview — Serverless features
- docs.novita.ai/api-reference/gpu-instance-create-instance.md — Billing modes (onDemand, monthly, spot), max 8 GPUs, VPC, network storage
- docs.novita.ai/api-reference/gpu-instance-list-products.md — Product fields incl. monthlyPrice, spotPrice, regions
- docs.novita.ai/guides/payment-methods — Stripe, minimum top-up, auto top-up
- docs.novita.ai/guides/faq — Payment methods (card, PayPal, no crypto), refunds, dedicated IP, restart behaviour
- docs.novita.ai/guides/quickstart — Sign-up methods, verification mention, Discord support
- novita.ai/user/register — Sign-up options, Sandbox credits
- docs.novita.ai/guides/gpu-instance-quickstart-connect-to-instance — SSH access
- docs.novita.ai/guides/gpu-instance-jupyterlab — Jupyter support
- docs.novita.ai/guides/gpu-instance-save-image — Image snapshots
- trust.novita.ai/ — SOC 2 Type II, GDPR, subprocessors
- trust.novita.ai/faq — DPA, SCCs, SOC 2 under NDA, CCPA
- status.novita.ai/ — Status page and published uptime
- novita.ai/legal/terms-of-service — Refund policy, governing law, export controls
- novita.ai/legal/privacy-policy — Entity details (none published), data transfers
AlternativesProviders people compare with it
DaytonaOperator · 8 GPU types · H100 $3.95not scored
ScalewayOperator · 5 GPU types · H100 $3.20Trust 85
LiumMarketplace · 84 GPU types · H100 $1.79Trust 40
EmpirioLabs AIOperator · 26 GPU types · H100 $3.80not scored
AceCloudOperator · 12 GPU typesTrust 73
RuncrateReseller · 20 GPU types · H100 $2.83not scoredQuestions about Novita AI
What GPUs does Novita AI offer?
Novita AI has 6 GPU types with a fresh offer today, for example H100 from $3.39, L40S from $0.55, RTX 5090 from $0.56, RTX 4090 from $0.33, RTX 6000 Ada from $0.77 per GPU-hour on-demand. The full list with prices and market medians is on this page.
How does Novita AI bill?
Billing granularity: per second. Stopped instance: Compute billing stops when the instance is stopped. Data of stopped instances is retained 7 days with storage fees of $0.005/GB/day, then deleted permanently. Pay-as-you-go instances release their hardware when stopped and may not restart if it has been reallocated.
How trustworthy is Novita AI?
Its Trust Score is 82/100 (v0.1), read from its own documents. Strong on infrastructure control, pricing & billing transparency. Weaker on company & track record. The score measures transparency and published guarantees, not customer satisfaction; every criterion and its source is listed above.
Where are Novita AI's data centres?
1 countries: —.
How do you pay Novita AI, and are there credits?
Payment by card. Free credits: Sign-up page: $100 in Sandbox credits valid 90 days (Agent Sandbox; not stated to apply to GPUs). FAQ mentions a $1 credit voucher revoked on suspicious sign-ups..