{ "as_of": "2026-09-14", "providers": [ { "provider": "CoreWeave", "uptime_sla_pct": ">=99.9% monthly uptime for instances deployed across multiple regions; >=99% for single-region/single-instance deployments (Terms of Service). Separate AI Object Storage SLA: >=99.9% monthly uptime (cwobject.com endpoint only, not cwlota.com).", "sla_measured_on": "Instance level: downtime = loss of external connectivity or persistent disk access for the Single Instance (single region) or all applicable running instances (multi-region), measured in minutes per month (partial minutes under one minute excluded). Storage SLA measured via error percentage (InternalError/ServiceUnavailable) over five-minute intervals.", "sla_remedy": "credits \u2014 tiered financial credits on the monthly bill. Compute multi-region: 10% (99-99.9%), 25% (95-98.999%), 50% (<94.999%); single region: 10% (95-99%), 25% (90-94.999%), 50% (<89.999%); capped at 50% of the month's amount due. Storage: 10%/25%/100% of affected zone's monthly bill. Credits apply to future use only (no refunds), must be claimed within 30 days, stated as sole and exclusive remedy.", "capacity_guarantee": "Yes, published in capacity-plans docs: 'Reserved Instances (RI) provide a guaranteed capacity reservation. During your term commitment, the reserved instances are available in your chosen region regardless of platform-wide demand.' Flex Reservations 'provide guaranteed peak capacity while reducing the cost of idle capacity.' On-Demand explicitly has 'no capacity guarantees' and 'may not be available' during peak demand.", "egress_fees": "Free \u2014 pricing page lists 'Data transfer between CoreWeave and internet', 'Data transfer within CoreWeave', and 'Egress and IOPS' all as Free. Zero Egress Migration (0EM) program covers egress costs from AWS/Azure/GCP when migrating in; AI Object Storage has no egress fees.", "billing_granularity": "Not explicitly published as per-second/minute/hour. Rates displayed per-hour; capacity-plans docs describe hourly charge calculation with 30-second usage-attribution intervals; ToS states monthly billing in arrears (invoices <$20,000 auto-charged to credit card, >=$20,000 via ACH/wire Net 15).", "self_serve": "no", "security_certs": "Trust page (coreweave.com/trust) displays SOC 2, ISO 27001, ISO 27017, ISO 27018 badges; other pages state SOC 2 Type II for Bare Metal and CoreWeave Kubernetes Service. GDPR/CCPA described only as 'designed to support customer compliance', not certified. No HIPAA claim found. Detailed reports under NDA via trust.coreweave.com (Conveyor).", "notes": "Sales-led onboarding: docs state 'When CoreWeave approves your organization, the CoreWeave Sales team sends an invitation email' \u2014 no public self-serve signup, though credit-card billing exists post-approval. SLA excludes pre-GA features, customer-caused issues, third-party problems, quota downtime, and pre-announced maintenance; maintenance policy also permits unplanned maintenance without notice. Reserved/committed pricing (up to 60% off on-demand) requires contacting sales.", "source_urls": [ "https://docs.coreweave.com/policies/terms-of-service", "https://docs.coreweave.com/docs/policies/terms-of-service/coreweave-ai-object-storage-policy", "https://docs.coreweave.com/platform/capacity-plans", "https://www.coreweave.com/pricing", "https://www.coreweave.com/trust", "https://docs.coreweave.com/get-started", "https://docs.coreweave.com/policies/terms-of-service/security-and-compliance", "https://www.coreweave.com/products/storage/zero-egress-migration-program" ] }, { "provider": "Lambda", "uptime_sla_pct": null, "sla_measured_on": null, "sla_remedy": "none stated \u2014 Terms of Service contain no uptime SLA or service credits; fees are explicitly 'non-cancelable' and 'non-refundable' except as expressly set forth. Only published SLA-like commitment is support response times for 1-Click Clusters: Severity 1 = 4-hour, Severity 2 = 8-hour, Severity 3 = 10-hour initial response, service requests = 24-hour.", "capacity_guarantee": "No published guarantee language. Pricing page describes on-demand as 'self-serve, first-come access' and says 'Contact us for reserved capacity at our lowest prices' \u2014 reserved capacity exists but its guarantee terms are not published. Marketing pages reference 'industry-leading SLAs' and 'SLA-backed job scheduling' (Managed Slurm) with no numbers.", "egress_fees": "Free \u2014 billing docs state 'you are not charged for ingress or egress.'", "billing_granularity": "per-minute \u2014 on-demand instances priced hourly but 'billed in one-minute increments'; billing runs from launch (after health checks pass) to termination. Filesystem storage billed per GiB/month in one-hour increments. Weekly invoices.", "self_serve": "yes", "security_certs": "Trust page (lambda.ai/trust): 'SOC 2 Type II attestation' and 'ISO 27001, 27017, 27701, and 22301 certified'; CCPA logo displayed; trust.lambda.ai lists SOC 2 Type 2, ISO/IEC 27001:2022, 27017:2015, 27701:2019, ISO 22301:2019. No HIPAA claim and no GDPR certification claim on the trust page.", "notes": "No published uptime percentage found anywhere on lambda.ai or docs.lambda.ai \u2014 'not published' is the finding for on-demand cloud; any uptime SLA for reserved/supercluster deals would be in unpublished contracts. Self-serve requires a major credit card (prepaid/debit rejected) with a $10 refundable pre-authorization; purchases supported from only 18 listed countries/territories. Formerly lambdalabs.com, now lambda.ai.", "source_urls": [ "https://lambda.ai/legal/terms-of-service", "https://docs.lambda.ai/public-cloud/billing/", "https://docs.lambda.ai/public-cloud/manage-billing/", "https://docs.lambda.ai/public-cloud/1-click-clusters/support/", "https://lambda.ai/trust", "https://lambda.ai/pricing", "https://lambda.ai/superclusters" ] }, { "provider": "Crusoe", "uptime_sla_pct": "99.5% Monthly Uptime Percentage for GPU-enabled virtual machines (also >=99.5% for Persistent/Shared Disks; >=99.9% for Container Registry and Managed Kubernetes control plane). Marketing pages cite '99.5% uptime'. SLA version 1.1, effective July 21, 2026.", "sla_measured_on": "A single GPU-enabled virtual machine, per calendar month ('Monthly Uptime Percentage and Financial Credit are determined on a calendar month basis per GPU-enabled virtual machine'). Downtime = unavailability e.g. loss of internal/external network connectivity or hardware failure; for hardware failure, downtime runs from failure until 'a new machine is made available'. Excludes customer-requested maintenance and planned or unplanned Crusoe maintenance. No cluster-level SLA published.", "sla_remedy": "credits \u2014 'Financial Credits' applied to future bills: 10% of monthly bill for the affected service if uptime 95.00%-99.49%, 25% if 90.00%-94.99%, 100% if <90.00%. Sole and exclusive remedy; must request within 30 days with log files/monitoring data; capped at 100% of the amount due for the affected service that month; applied within 60 days.", "capacity_guarantee": "Reservations = 'an active reserved instance agreement, typically for a fixed quantity of a specific instance type (e.g. 128 A100-80GB GPUs) valid for a defined period of time'; reserved capacity auto-applied to matching running instances, unused capacity does not roll over, overage bills at on-demand. ToS 'Committed Instances': customer pays for the Commitment Term 'whether or not they are used'; purchases 'may not be cancelled or refunded'. No explicit published language guaranteeing physical capacity availability beyond the reserved instance agreement itself; reservations are sales-negotiated.", "egress_fees": "Free \u2014 'At this time, Crusoe Cloud does not charge for network ingress or egress, either within a VPC or to/from the public internet' (docs and pricing page); marketing says 'zero egress fees'.", "billing_granularity": "Pricing page states on-demand is 'billed per-hour (or per-second)'; prices listed per GPU-hr. Exact metering increment not unambiguously published; billing dashboard updates daily.", "self_serve": "yes", "security_certs": "Cloud page claims 'ISO 27001, ISO 42001, SOC 2 Type I and II certified and GDPR-compliant standards'; blog posts confirm SOC 2 Type II attestation and ISO 27001 + ISO 42001 certifications; Trust Center at trust.crusoe.ai (JS-gated, contents not fetchable). No HIPAA claim found on Crusoe Cloud pages.", "notes": "Self-serve signup via console (Google/GitHub/email), then enable billing with 'a valid, non-prepaid credit card'; $5 free credits for serverless inference/fine-tuning. Reservations/committed pricing and newest GPUs (GB200, B200, MI355X) require contacting sales. Full SLA text is published at legal.crusoe.ai but the page is a JS single-page app; SLA content extracted from raw page payload (v1.1 effective Jul 21, 2026; prior version Dec 23, 2025). SLA exclusions notably include both planned AND unplanned Crusoe maintenance per Technical Support Services guidelines, plus pre-GA features, factors outside Crusoe's control, customer/third-party hardware-software, quota limits.", "source_urls": [ "https://legal.crusoe.ai/", "https://www.crusoe.ai/cloud", "https://www.crusoe.ai/cloud/pricing", "https://www.crusoe.ai/", "https://docs.crusoecloud.com/networking/overview/index.html", "https://docs.crusoecloud.com/create-an-account", "https://docs.crusoecloud.com/usage-billing/reservations/overview", "https://docs.crusoecloud.com/usage-billing/reservations/managing-reservations/index.html", "https://www.crusoe.ai/resources/blog/crusoe-cloud-achieves-iso-27001-and-42001-certifications", "https://trust.crusoe.ai/" ] }, { "provider": "Nebius", "uptime_sla_pct": "99.50% Service Uptime for Compute Cloud (single VM). Other published levels per service: Managed Kubernetes, PostgreSQL, MLflow, Object Storage, VPC each have their own service-level pages; Standalone Applications is 95.00%. No GPU-cluster-specific SLA published \u2014 compute SLA has no GPU-specific service level terms.", "sla_measured_on": "Single Virtual Machine \u2014 unavailability defined as 'Loss of external connectiveness or a boot disk of a Virtual Machine'; separate tier for non-replicated disk placement groups (two or more disks in error state within an 8-hour interval). Measured per Reporting Period; service levels can vary by region. DoS/DDoS on the customer's VM is excluded.", "sla_remedy": "credits \u2014 compensation as a discount against service fees in the reporting period: 10.0% of the service fee if uptime is 99.50%-99.00%, 15.0% if 99.00%-95.00%, 30.0% if below 95.00%. 'Compensation is the sole and exclusive remedy'; credits cannot exceed total cost of service usage in the period, apply only as discounts to fees, no cash refunds.", "capacity_guarantee": "Not published as a guarantee. Reserved/committed capacity ('Pay up to 35% less than on-demand rates by reserving large-scale clusters for multiple months') is sales-negotiated via addendum; docs state 'Extending the discounts is subject to the available capacity at the time of the request' and cancellation only 'if allowed by the terms of your addendum'. No standard published capacity-guarantee terms.", "egress_fees": "No general compute/VM egress charge published: 'The Virtual Networks service in Nebius AI Cloud is provided free of charge' (VPC pricing page lists no traffic pricing). Object Storage egress IS charged: $0.0150 per GiB (standard); Intelligent tier egress 'Free*' limited to 3x the customer's monthly average stored capacity; Enhanced storage egress free.", "billing_granularity": "per-second \u2014 docs state 'Billing unit: 1 second' with 'Pricing unit: 1 hour (3600 seconds)' for GPUs, vCPUs and RAM; charges below one hour prorated; 'Computing resources of stopped VMs are not charged (this does not apply to storage volumes)'. Serverless AI also bills per-second.", "self_serve": "yes", "security_certs": "Trust Center lists: SOC 2 Type II 'With HIPAA', SOC 3, ISO 27001, ISO 27018, ISO 27032, ISO 27701, ISO 27799, ISO 22301, ISO 42001 (AI Management Systems), CSA STAR Level 1, alignment with NIS 2 and DORA, and 'GDPR and CCPA-compliant policies and practices'.", "notes": "Self-serve: individuals (bank card only) and companies (bank card except Israel, or bank transfer); activation 'Automatic after adding a card and paying the initial $25'; only Israel-based companies require manual activation. On-demand GPUs (e.g. H100 $3.85/GPU-hr) available without sales; committed discounts and GB200/GB300 NVL72 require sales. Umbrella SLA at docs.nebius.com/legal/sla defers all percentages to per-service pages; SLA applies only to fee-paying usage and excludes customer actions, third-party software, preventive maintenance, emergency works, government requests, AUP violations, quota breaches, force majeure. Compensation must be claimed and is capped at the period's spend. Compute SLA covers single VMs only; no published cluster-wide or GPU-health SLA.", "source_urls": [ "https://docs.nebius.com/legal/sla", "https://docs.nebius.com/legal/sla-levels", "https://docs.nebius.com/legal/sla-levels/compute", "https://nebius.com/prices", "https://docs.nebius.com/compute/resources/pricing", "https://docs.nebius.com/vpc/resources/pricing", "https://docs.nebius.com/signup-billing/payments/payers", "https://docs.nebius.com/signup-billing/billing-models/committed-usage", "https://nebius.com/trust-center" ] }, { "provider": "Voltage Park", "uptime_sla_pct": "99.5% monthly uptime ('>= 99.5%') \u2014 same target for both the VM SLA and the Node SLA", "sla_measured_on": "Per single GPU-enabled virtual machine (VM SLA) and per single bare-metal node (Node SLA), tracked per calendar month; the Computer Services Agreement applies the VM SLA to virtual machine orders and the Node SLA to bare metal orders", "sla_remedy": "credits \u2014 tiered Financial Credits: 10% of monthly bill for 95.00%-<99.50% uptime, 25% for 90.00%-<95.00%, 100% for <90.00%; capped at that service's monthly charges; customer must claim within 30 days with logs/monitoring data; credits applied to future use within 60 days of request", "capacity_guarantee": null, "egress_fees": "Free \u2014 pricing page states 'No hidden ingress, egress or support costs'; no separate egress price list published", "billing_granularity": "per-hour rates \u2014 Computer Services Agreement states 'Fees per hour for the Services' charged 'on a continuous and incremental basis' and notes 'charges may be more frequent than once per hour'; no per-second/per-minute granularity published", "self_serve": "yes", "security_certs": "SOC 2 Type II, ISO/IEC 27001, and HIPAA 'for eligible workloads' per voltagepark.com/security ('compliant with recognized industry standards, including ISO / IEC 27001, SOC 2 Type II, and HIPAA for eligible workloads'); Tier 3+ data centers; full controls at trust.voltagepark.com", "notes": "Self-serve: pricing page says 'Self-serve. Up within 15 mins. No minimum terms' for on-demand (H100 from $1.99/hr) with dashboard deployment and self-serve fund deposits; Long-Term Reserve (6+ month contracts, 32-8,000+ H100s) requires contacting sales. SLA exclusions: pre-GA features, scheduled/communicated maintenance, factors outside Voltage Park's control, customer/third-party systems, agreement violations. No capacity-guarantee language found in the SLA or Computer Services Agreement \u2014 'Long-Term Reserve' / 'Dedicated Reserve' products exist but no published guarantee of reserved-capacity access. trust.voltagepark.com renders client-side and could not be scraped, so certs verified from the /security page. Docs billing page notes stopped instances are still charged at the full running rate.", "source_urls": [ "https://www.voltagepark.com/virtual-machine-sla", "https://www.voltagepark.com/computer-services-agreement", "https://www.voltagepark.com/pricing", "https://www.voltagepark.com/security", "https://docs.voltagepark.com/organization/billing", "https://trust.voltagepark.com/" ] }, { "provider": "Together AI", "uptime_sla_pct": "GPU Clusters: none published \u2014 Terms of Service state 'No guarantees are made with respect to the Services' quality, stability, uptime, or reliability, unless otherwise agreed between the parties in an order form.' Inference only: Provisioned Throughput page publishes 'Together targets a 99% monthly success rate for eligible requests' (marketed as '99% uptime SLA'); official blog publishes contractual 99% (single-DC) / 99.9% (multi-DC) uptime SLAs for dedicated inference deployments", "sla_measured_on": "Inference only: eligible requests to the contracted model through the PTU endpoint within purchased capacity (monthly success rate); blog states uptime is 'measured at inference completion, not load balancer.' No SLA measurement basis published for GPU clusters (node/cluster)", "sla_remedy": "none stated \u2014 no credit/refund/remedy terms published on provider pages for either PTU or GPU clusters; ToS says fees are non-refundable; any SLA remedies live in unpublished order forms", "capacity_guarantee": "PTU: 'Reserved capacity means your traffic doesn't compete... Guaranteed tokens per minute' (send up to contracted normalized TPM; overage falls back to serverless best-effort with no availability guarantee). GPU Clusters: reserved capacity 'Locked in for your duration' (terms up to 6 months, up to 4,000+ GPUs reserved vs up to 256 GPUs on-demand); reserved credits charged upfront for the full duration once provisioned", "egress_fees": "Free \u2014 docs: 'Together AI does not charge for data transfer into or out of your cluster... You pay only for compute and storage'; Instant Clusters GA post: 'Data transfer: free egress and ingress'", "billing_granularity": "per-hour \u2014 GPU clusters priced and billed per GPU per hour (on-demand: pay only for time cluster is running; commitment options Hourly, 1-6 Days, 1 Week-3 Months; reserved deducted upfront for full duration); no per-second/per-minute billing published", "self_serve": "partial", "security_certs": "SOC 2 Type II and ISO 27001:2022 (badges on GPU Clusters page; SOC 2 Type 2 report announced on official blog, 'officially ISO 27001:2022 certified'); HIPAA compliance claimed (encryption in transit/at rest, audit logging, BAAs); GDPR referenced as supported; trust center at trust.together.ai", "notes": "Self-serve is yes for Instant Clusters (console/CLI/API/Terraform, 'Spin up a cluster in minutes', prepaid credit-card balance per ToS) but on-demand is capped at ~256 GPUs; larger/longer reserved capacity and dedicated inference require contacting sales \u2014 recorded as partial. Key nuance: Together's published SLA numbers (99% PTU success rate; 99%/99.9% single/multi-DC in blog) apply to INFERENCE offerings, not GPU cluster/dedicated compute rentals \u2014 for GPU clusters no uptime SLA is published and the ToS explicitly disclaims uptime guarantees absent an order form. PTU SLA excludes overage traffic and has a one-month minimum term. Support page offers tiered support (Silver/Gold) with response-time SLAs, not uptime SLAs. The '99.9% SLA enterprise tier with automatic service credits' claim circulating in third-party sources was not found on any together.ai page and is excluded. Instant Clusters storage $0.16/GiB-month.", "source_urls": [ "https://www.together.ai/terms-of-service", "https://www.together.ai/pricing", "https://www.together.ai/gpu-clusters", "https://www.together.ai/provisioned-throughput", "https://docs.together.ai/docs/gpu-clusters-billing", "https://www.together.ai/blog/99-9-uptime-for-inference", "https://www.together.ai/blog/together-instant-clusters-ga", "https://trust.together.ai/" ] }, { "provider": "FluidStack", "uptime_sla_pct": "99% network availability per calendar month (current Terms & Conditions); a legacy standalone SLA PDF dated 14 July 2023 states a target service level of '[99.5]% in respect of the cloud infrastructure' per calendar month (figure appears in brackets, i.e. as a negotiable placeholder)", "sla_measured_on": "network \u2014 'the network is deemed available if the border routers and core-switches are responding to Fluidstack's monitoring tools', polled at least every 15 seconds (per T&C); legacy SLA PDF measures 'cloud infrastructure' uptime and adds support response/resolution targets (Critical: 1hr response / 6hr resolution)", "sla_remedy": "credits \u2014 'Service credits are calculated on a monthly basis and are Customer's sole and exclusive remedy for Downtime Periods each month'; explicitly no refunds: 'Fluidstack will not in any circumstances be obliged to pay any money or make any refund to the Customer in respect of any Downtime Periods'; credit amounts are set per Order, not published", "capacity_guarantee": null, "egress_fees": null, "billing_granularity": "not published as a rate card; T&C contemplates fees 'charged on a per-minute or per-hour basis' with prepaid deposit possibly required, plus subscription billing with auto-renewal", "self_serve": "no", "security_certs": null, "notes": "capacity_guarantee: nothing published about guaranteed reserved-capacity access. egress_fees: not mentioned in T&C, legal index, or docs. self_serve: no public pricing page (fluidstack.io/pricing is 404) and homepage has no product signup CTA (mission/recruiting focused, 'scale to 100s of GW'); T&C references orders 'placed electronically via the Fluidstack Console' but there is no public self-serve credit-card path on the current site \u2014 engagement is sales/contract-led. security_certs: no trust/security page found (fluidstack.io/resources/security 404); the legal index (terms, AUP, privacy, DPA, non-circumvention) contains no cert claims; job postings reference SOC 2 / ISO 27001 / FedRAMP compliance programs but that is not a published certification claim, so recorded as null. The 2023 SLA PDF is hosted on their webflow CDN and is not linked from the current /legal index, so the T&C availability clause (99%, credits-only) should be treated as the current published position.", "source_urls": [ "https://fluidstack.io/legal/terms-conditions", "https://cdn.prod.website-files.com/64e49a6c77bc12449a05e6a2/6502005b4edf9b61ccbf35d1_SLA.pdf", "https://fluidstack.io/legal", "https://fluidstack.io/", "https://docs.fluidstack.io", "https://fluidstack.io/pricing (404)", "https://fluidstack.io/resources/security (404)" ] }, { "provider": "Verda", "uptime_sla_pct": null, "sla_measured_on": null, "sla_remedy": "none stated in published terms \u2014 Terms & Conditions provide the service 'as is' and disclaim that 'the operation or availability of the Service will be uninterrupted or error-free', with no credit/refund schedule; marketing pages promise 'sensible SLAs and fair compensation for service disruptions' but publish no specifics", "capacity_guarantee": null, "egress_fees": "partially published: container registry page states 'No egress fees to Verda compute'; no general instance/network egress or data-transfer price list published on pricing page, T&C, or docs", "billing_granularity": "prepaid 10-minute increments for Pay As You Go \u2014 docs state 'billing calculated in pre-paid 10-minute increments'; prices displayed as hourly rates; 'if a resource is terminated before the billed period ends, the unused portion is automatically refunded within the next billing period'", "self_serve": "yes", "security_certs": "SOC 2 Type II ('Verda is SOC 2 Type II compliant' banner + blog confirming completed Type II audit covering Security, Availability, Confidentiality for the 2025 audit period, report available under NDA); ISO 27001, 27017, 27018, 27701; C5; GDPR compliant; audits published at trust.verda.com (Verda Trust Center, content is JS-rendered)", "notes": "Formerly DataCrunch (datacrunch.io); rebranded to Verda \u2014 datacrunch.io content has migrated to verda.com/docs.verda.com. uptime_sla_pct is null because no contractual SLA percentage is published anywhere found: homepage claims 'Historical uptime of over 99.9%' and 'over 99.95% uptime' (NVIDIA Preferred Partner context), but these are historical marketing claims, not commitments, and the T&C expressly disclaims uninterrupted operation. capacity_guarantee: reserved pricing exists (1-24 month terms, discounts 2% at 1mo / 3% at 3mo / 4% at 6mo / 8% at 1yr / 25% at 2yr per pricing page) but no published guarantee-of-access language for reserved capacity. Spot listed at 50% of on-demand. self_serve: 'Instant, self-service access to the latest compute on the market at any scale \u2014 no negotiations, no contracts' with console signup; docs list credit card (one-time and auto top-up), SEPA, and invoice payment. Governed by Finnish law, arbitration in Helsinki.", "source_urls": [ "https://verda.com/", "https://verda.com/terms-and-conditions", "https://verda.com/pricing", "https://docs.verda.com/welcome-to-verda/pricing-and-billing/", "https://verda.com/container-registry", "https://verda.com/ai-cloud", "https://verda.com/blog/soc-2-type-ii", "https://trust.verda.com", "https://docs.verda.com/resources/services-overview/" ] }, { "provider": "AWS", "uptime_sla_pct": "99.99% region-level; 99.5% instance-level", "sla_measured_on": "Amazon Compute SLA: 99.99% applies when instances run concurrently across 2+ Availability Zones in a region; 99.5% applies to each individual EC2 instance. One claim type per instance, credits cannot be stacked. SLA text does not carve out GPU instances \u2014 applies to EC2 generally.", "sla_remedy": "credits \u2014 10% credit for uptime 99.0-99.99%, 30% for 95.0-98.99%, 100% below 95.0% (same tiers for both SLAs); instance-level SLA adds no charge for an instance unavailable >6 min in an hour. Claim-based, within two billing cycles, credits apply only to future charges.", "capacity_guarantee": "No published 'guarantee' wording. EC2 Capacity Blocks for ML lets you 'reserve' GPU instances (P6e-GB200/P6-B300/P6-B200/P5en/P5e/P5/P4d/Trn2/Trn1) up to 8 weeks ahead, up to 6 months, 1-64 instances (512 GPUs), dynamic supply/demand pricing. On-Demand Capacity Reservations 'help ensure that you always have access' \u2014 'help ensure', not 'guarantee'; future-dated Capacity Reservations exclude the P family (Capacity Blocks is the P-family path).", "egress_fees": "100 GB/month data transfer out to internet free (aggregated across AWS services/regions, excl. China/GovCloud); beyond that tiered per-GB, first tier $0.09/GB confirmed on AWS-controlled pages. Full tier table is JS-rendered on the pricing page and could not be captured directly.", "billing_granularity": "per-second with 60-second minimum for Linux (incl. GPU P-family); Capacity Blocks priced separately (dynamic, reservation-based)", "self_serve": "partial \u2014 credit-card signup yes, but published default quota for 'Running On-Demand P instances' is 0 vCPUs (also 0 for G/VT, Inf, Trn, DL, F); GPU launch requires a Service Quotas increase request first", "security_certs": "SOC 2, ISO 27001, HIPAA (via BAA), FedRAMP \u2014 all listed on aws.amazon.com/compliance/programs/", "notes": "AWS never publishes the word 'guarantee' for GPU capacity. Egress tier table ($0.09/$0.085/$0.07/$0.05 per GB commonly cited) only partially verifiable on provider-controlled pages due to JS rendering; 100 GB free tier and $0.09/GB first tier confirmed.", "source_urls": [ "https://aws.amazon.com/compute/sla/", "https://aws.amazon.com/ec2/capacityblocks/", "https://aws.amazon.com/ec2/pricing/on-demand/", "https://aws.amazon.com/compliance/programs/", "https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/ec2-capacity-reservations.html", "https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/ec2-on-demand-instances.html", "https://docs.aws.amazon.com/solutions/latest/data-transfer-hub/cost.html" ] }, { "provider": "Microsoft Azure", "uptime_sla_pct": "99.99% (2+ VMs across 2+ Availability Zones); 99.95% (2+ VMs in Availability Set / Dedicated Host Group); single instance: 99.9% (Premium SSD/Ultra Disk), 99.5% (Standard SSD), 95% (Standard HDD)", "sla_measured_on": "'Virtual Machine Connectivity' \u2014 connectivity to at least one instance for multi-instance deployments; single-instance tier depends on disk type for all OS and data disks. No GPU-specific SLA; standard VM SLA applies to ND/NC-series by deployment configuration.", "sla_remedy": "credits \u2014 multi-instance: <99.99%/<99.95% = 10%, <99% = 25%, <95% = 100%; single-instance tiers step down similarly by disk type. 'Service Credits are Customer's sole and exclusive remedy.' Claim-based.", "capacity_guarantee": "Explicitly published but NOT for flagship AI GPUs: On-demand Capacity Reservations 'Provides capacity guarantee in the specified location' with an SLA, but docs state 'Support for other VM series isn't currently available: GPU-accelerated compute: ND-series, NCads H100 v5-series.' Supported GPU series limited to NCv3/NCasT4_v3/NC_A100_v4/NV-series and similar. Also published: 'A quota is a credit limit on Azure resources, not a capacity guarantee.'", "egress_fees": "First 100 GB/month free; internet egress from North America/Europe: next 10 TB $0.08/GB, next 40 TB $0.065, next 100 TB $0.06, next 350 TB $0.04 (higher rates from Asia/Australia/MEA and South America); >500 TB/month contact Microsoft", "billing_granularity": "per-minute \u2014 'We charge for the number of full minutes your virtual machine is running'; no sub-minute billing published for VMs", "self_serve": "partial \u2014 pay-as-you-go credit-card signup, but GPU families (NCasT4_v3, NC_A100_v4, NDv2, etc.) 'start with a default of zero cores'; quota increase via portal, may escalate to a support request; Free Trial/Students subscriptions cannot get GPU quota", "security_certs": "SOC 1/2/3, ISO 27001/27017/27018/27701, HIPAA, FedRAMP \u2014 listed on Azure compliance documentation hub (~100 offerings)", "notes": "Canonical global VM SLA text now lives in the downloadable 'SLA for Online Services' licensing document; exact figures verified via Microsoft's mirrored Azure China SLA page plus Microsoft Learn. Key gap: ND-series/H100-class excluded from the published capacity guarantee.", "source_urls": [ "https://www.microsoft.com/licensing/docs/view/Service-Level-Agreements-SLA-for-Online-Services", "https://www.azure.cn/en-us/support/sla/virtual-machines/", "https://learn.microsoft.com/en-us/azure/reliability/reliability-virtual-machines", "https://azure.microsoft.com/en-us/pricing/details/bandwidth/", "https://azure.microsoft.com/en-us/pricing/details/virtual-machines/linux/", "https://learn.microsoft.com/en-us/azure/virtual-machines/capacity-reservation-overview", "https://learn.microsoft.com/en-us/azure/quotas/per-vm-quota-requests", "https://learn.microsoft.com/en-us/azure/machine-learning/how-to-manage-quotas", "https://learn.microsoft.com/en-us/azure/compliance/" ] }, { "provider": "Google Cloud", "uptime_sla_pct": ">=99.99% multi-zone; >=99.95% single instance Memory Optimized; >=99.9% single instance all other machine families (GPU A3/A2/G2 fall here); >=99.9% multi-zone on Standard network tier", "sla_measured_on": "Monthly Uptime Percentage per calendar month, per Project per Region (multi-zone) or per instance. Downtime = loss of external connectivity or persistent disk access in periods of >=1 consecutive minute. Spot VMs (the discounted GPU path) are explicitly excluded from the Compute Engine SLA.", "sla_remedy": "credits \u2014 Financial Credits on future bills, sole and exclusive remedy, request within 60 days. Multi-zone: 99.00-<99.99% = 10%, 95.00-<99.00% = 25%, <95% = 100%. Single instance (other families): 95.00-<99.90% = 10%, 90.00-<95.00% = 25%, <90% = 100%.", "capacity_guarantee": "Published AI Hypercomputer consumption options: standard reservations ('very high level of assurance for capacity'); future reservations for capacity blocks (1-year+ commitment for A4X Max/A4X/A4/A3-class, quota auto-increased); Dynamic Workload Scheduler calendar mode = reserved exclusive-access dense GPU capacity up to 90 days, no commitment, up to 53% discount; DWS flex-start = best-effort, not guaranteed. No financial remedy published for reservation non-delivery.", "egress_fees": "Premium Tier internet egress from US: ~1 GiB free, then $0.12/GiB (to 1 TiB), $0.11 (1-10 TiB), $0.08-0.085 above 10 TiB destination-dependent. Standard Tier: first 200 GiB/month free, then $0.085 (to 10 TiB), $0.065 (10-150 TiB), $0.045 (>150 TiB). Inbound free.", "billing_granularity": "per-second with 1-minute minimum \u2014 billing model page states this applies to 'all vCPUs, GPUs, and GB of memory'", "self_serve": "partial \u2014 credit-card signup works, but GPU quota requires an approved quota increase request (per-model/per-region quotas); Free Trial billing accounts cannot request quota changes at all, so no GPUs on free trial", "security_certs": "SOC 2 (SSAE 18 reports via Compliance Reports Manager), ISO 27001 (Compute Engine in scope), HIPAA (BAA supported), FedRAMP High P-ATO (requires Assured Workloads for FedRAMP High)", "notes": "SLA text has no GPU-specific exclusion but defers to per-product documentation exclusions; Spot/preemptible GPUs carry no SLA. DWS/capacity-block reservations are capacity commitments without a published uptime-SLA-style credit remedy. Mexico and Stockholm regions have lower multi-zone SLO (99.95%).", "source_urls": [ "https://cloud.google.com/compute/sla", "https://cloud.google.com/vpc/network-pricing", "https://cloud.google.com/compute/vm-instance-pricing", "https://cloud.google.com/compute/resource-usage", "https://cloud.google.com/compute/docs/instances/spot", "https://cloud.google.com/ai-hypercomputer/docs/consumption-models", "https://cloud.google.com/compute/docs/instances/future-reservations-overview", "https://cloud.google.com/security/compliance/soc-2", "https://cloud.google.com/security/compliance/iso-27001", "https://cloud.google.com/security/compliance/hipaa", "https://cloud.google.com/security/compliance/fedramp" ] }, { "provider": "Oracle Cloud/OCI", "uptime_sla_pct": ">=99.9% Monthly Uptime Service Commitment for OCI Compute (Availability SLA); credit thresholds keyed at 99.99% multi-AD, 99.95% single-AD, 99.9% single instance; separate Manageability (control plane) and Performance SLAs each >=99.9%", "sla_measured_on": "Per calendar month; 'Unavailable' = problem preventing external connectivity with all instances across multiple Availability Domains (multi-AD), all instances across multiple Fault Domains (single-AD region), or each single instance. GPU shapes (H100/H200/B200/GB200/GB300/B300/A100/A10/L40S/MI300X/MI355) explicitly listed in the Compute SLA table. Source: PaaS/IaaS Pillar Document, July 2026 edition, section 3.6.21.", "sla_remedy": "credits \u2014 exclusive remedy, claim-based. Multi-AD: <99.99% to >=99.0% = 10%, <99.0% to >=95.0% = 25%, <95% = 100%. Single-AD: <99.95% to >=99.0% = 10%, <99.0% to >=95.0% = 25%, <95% = 100%. Single instance: <99.9% = 100% credit.", "capacity_guarantee": "Capacity reservations published: 'Assurance that you have the capacity necessary... available for your tenancy to consume at any time. No size or time commitments'; unused reserved capacity billed at 85% of shape rate. The capacity-reservation doc does not state GPU shape eligibility (GPU shapes not mentioned there); not available on Free Tier. No GPU calendar/scheduled-capacity product equivalent to GCP DWS published.", "egress_fees": "First 10 TB/month outbound free (all listed origin regions); over 10 TB: $0.0085/GB (NA/Europe/UK origin), $0.025/GB (APAC/Japan/South America), $0.05/GB (Middle East/Africa), per Oracle's price-list page and its backing API. Inbound free.", "billing_granularity": "per-second; VM instances 1-minute minimum, bare metal 1-hour minimum \u2014 most current OCI GPU shapes are bare metal (BM.GPU.*), so the 1-hour minimum applies; GPU SKUs priced as 'GPU Per Hour' unit rates", "self_serve": "partial \u2014 Pay As You Go signup is self-serve, but the published service-limits table lists 'Contact Us' for all GPU shape families (gpu3-count, gpu4-count, gpu-a10-count, gpu-a100-v2-count, etc.) under both Universal Credits and PAYG/Trial \u2014 GPU access requires contacting Oracle for a limit increase", "security_certs": "SOC 1/2, ISO 27001, HIPAA attestations, FedRAMP, DoD SRG, CSA STAR, PCI DSS \u2014 listed on Oracle cloud compliance page with per-service applicability in Oracle's attestation directory", "notes": "Headline commitment is 99.9% monthly uptime; the 99.99% figure appears only as the top credit threshold for multi-AD deployments, though Oracle markets 'end-to-end SLAs covering performance, availability, and manageability'. Newer H100/H200-era limit names not visible in fetched service-limits text, but all listed GPU families show 'Contact Us' \u2014 treat GPU quota as sales-mediated.", "source_urls": [ "https://www.oracle.com/cloud/sla/", "https://www.oracle.com/contracts/docs/paas_iaas_pub_cld_srvs_pillar_4021422.pdf", "https://www.oracle.com/cloud/price-list/#pricing-networking", "https://apexapps.oracle.com/pls/apex/cetools/api/v1/products/?currencyCode=USD", "https://docs.oracle.com/en-us/iaas/releasenotes/changes/20facdce-34bf-4648-a0a7-de5db73657ce/", "https://docs.oracle.com/en-us/iaas/Content/General/Concepts/servicelimits.htm", "https://docs.oracle.com/en-us/iaas/Content/Compute/Tasks/reserve-capacity.htm", "https://www.oracle.com/corporate/cloud-compliance/" ] }, { "provider": "Hyperstack", "uptime_sla_pct": "99.5% minimum uptime \u2014 but the same SLA page also states 'minimum uptime of 100.0%' in another occurrence; published commitment is internally inconsistent", "sla_measured_on": "'The Services' broadly \u2014 appendix lists GPU Processing, CPU Processing, Memory, Storage Capacity. Calculated monthly by NexGen 'in our sole discretion by reference to our available monitoring data' as (Total Time - Downtime)/Total Time; excludes scheduled maintenance, force majeure, customer-caused interference.", "sla_remedy": "credits \u2014 pro-rata rebate to account for fees paid in the affected billing cycle, capped at 100% of the amount paid for the affected Service(s); no graduated tiers. Claim must be emailed within 5 calendar days of end of the monthly billing cycle.", "capacity_guarantee": "Published Reservation Pricing on the GPU pricing page (per-model reserve options, 'Starting from' rates, invoiced monthly) plus a dedicated H100 SXM reservation page; 'Contact us directly for longer-term contracts' for large reserved deals", "egress_fees": "Free \u2014 pricing page: 'Egress/Ingress traffic: Free'; FAQ: 'No, there are no additional charges for bandwidth or data transfer.'", "billing_granularity": "per-minute for on-demand; reservations invoiced monthly; no minimum stated", "self_serve": "yes \u2014 credit card via Stripe ('add your credit card details and you're set!'); sales contact only for longer-term reservation contracts", "security_certs": "SOC 2 Type 2 attestation (July 2026, auditor Insight Assurance, per NexGen Cloud announcement \u2014 tied to 'Secure Private Cloud', on-demand scope not explicitly confirmed); ISO 27001 in progress, not achieved; no HIPAA/FedRAMP claims found", "notes": "SLA addendum's conflicting 99.5% vs 100.0% figures are both live on the same page \u2014 treat the commitment as ambiguous. nexgencloud.com/security page is stale (still says SOC 2 Type 1 expected Q2 2025), contradicted by the July 2026 Type 2 announcement. 5-day claim window is unusually short. SLA not tiered by service (no VM vs API vs network distinction).", "source_urls": [ "https://www.hyperstack.cloud/slas", "https://www.hyperstack.cloud/gpu-pricing", "https://www.nexgencloud.com/security", "https://www.nexgencloud.com/news/nexgen-cloud-achieves-soc-2-type-2-attestation-strengthening-verified-security-assurance-for-enterprise-ai" ] }, { "provider": "RunPod", "uptime_sla_pct": "No public SLA for self-serve tiers; '99.99% uptime SLA' appears only as enterprise-agreement marketing copy. ToS: 'Runpod does not make any specific uptime warranties with respect to the Community Cloud Offerings.'", "sla_measured_on": null, "sla_remedy": "none stated publicly \u2014 ToS: 'all Fees for the Service are non-refundable' except as set forth in the Terms; no service-credit schedule published; enterprise SLA remedies are contractual and unpublished", "capacity_guarantee": "Enterprise/sales-led only: 'Reserved Clusters: Dedicated GPU clusters with guaranteed availability, custom configurations, SLA-backed uptime'; 'Guaranteed capacity sized against your forecast, priced for commitment'. Self-serve Savings Plans (3/6 months prepaid) cover GPU compute costs only, no capacity guarantee wording.", "egress_fees": "Free \u2014 docs state Pods have 'no fees for data ingress or egress'", "billing_granularity": "per-second \u2014 Pods 'billed by the second for compute and storage' (network volumes billed hourly); no minimum stated", "self_serve": "partial \u2014 Pods/Serverless fully self-serve from console; reserved capacity and Reserved Clusters require sales ('Speak with our team to reserve')", "security_certs": "ISO/IEC 27001:2022 (cert RUN-ISMS-20260827, valid 2026-08-27 to 2029-08-26); SOC 2 Type II (Sensiba LLP); SOC 3; 'maintains HIPAA and GDPR programs'. Secure Cloud runs on 'vetted infrastructure partners who meet enterprise security standards including SOC 2, ISO 27001, and PCI DSS' \u2014 partner-DC claims apply to Secure Cloud, not Community Cloud.", "notes": "No measurement definition (VM vs API vs network) published anywhere for the 99.99% enterprise figure. Compliance page caveat: 'Compliance coverage can vary by workload, region, provider, and deployment model.' Secure Cloud vs Community Cloud priced differently (e.g., H100 PCIe $2.89/hr vs $1.99/hr); Community Cloud is peer-hosted with explicit no-uptime-warranty language.", "source_urls": [ "https://www.runpod.io/legal/terms-of-service", "https://www.runpod.io/legal/compliance", "https://www.runpod.io/enterprise", "https://www.runpod.io/pricing", "https://docs.runpod.io/pods/pricing", "https://docs.runpod.io/references/security-and-compliance" ] }, { "provider": "Vast.ai", "uptime_sla_pct": null, "sla_measured_on": null, "sla_remedy": "none stated \u2014 ToS affirmatively disclaims availability: 'Company cannot guarantee the Website and that Company Services will be always available' and 'Company shall have no liability whatsoever for any loss...during any downtime'. Credits spent are final and non-refundable; unspent balance refunds at sole discretion.", "capacity_guarantee": "On-demand instances 'Cannot be interrupted' (host-set fixed price); interruptible instances are bid-based and 'Can be stopped by higher bids'; reserved instances offer prepaid discounts (up to 50%) with 'Same priority as on-demand'. No published guarantee that on-demand capacity will be available to start.", "egress_fees": "Host-set per-host bandwidth pricing, shown during instance selection: 'Data transfer costs vary by host and include both upload and download traffic. Charges apply per byte transferred' \u2014 charged 'regardless of what state [the instance] is in'. No free-egress claim.", "billing_granularity": "per-second for active rental ('charged...for every second your instance is in the active/connected state'); storage billed continuously including while stopped", "self_serve": "yes \u2014 credit cards via Stripe, plus crypto (BitPay, Crypto.com); no sales contact needed for on-demand", "security_certs": "SOC 2 Type II certified (platform-level); SOC 3 report available on request; HIPAA workloads supported on Secure Cloud tier with BAAs 'available for qualifying customers'. Docs say Secure Cloud providers are vetted datacenters with ISO 27001 / Tier 3-4 standards, but compliance page caveats partner certs 'are encouraged...but are not strictly required'.", "notes": "Marketplace: no uptime SLA anywhere; the substitute is (1) per-second billing only while active and (2) host 'reliability scores' \u2014 verification requires reliability over 90% and unstable machines are auto-deverified, but this is a historical score, not a contractual commitment. Tier distinction: 'Secure Cloud' = certified/vetted datacenter partners (Tier 3/4, HIPAA-eligible) vs community/unverified hosts where 'Provider security varies significantly'.", "source_urls": [ "https://vast.ai/terms", "https://vast.ai/compliance", "https://docs.vast.ai/guides/reference/faq/billing.md", "https://docs.vast.ai/guides/reference/faq/rental-types.md", "https://docs.vast.ai/guides/reference/faq/security.md", "https://docs.vast.ai/guides/pricing.md", "https://docs.vast.ai/host/verification-stages.md" ] }, { "provider": "Scaleway", "uptime_sla_pct": "99.5% SLO for GPU Instances (tiers: Shared Instances 99%, Dedicated 99.5%, Specialized 99.5%, GPU 99.5%; some legacy families have no SLO)", "sla_measured_on": "External connectivity of the instance \u2014 'Unavailability Period' = loss of external connectivity for at least 4 continuous minutes, measured as monthly availability rate (MAR); excludes components under client control", "sla_remedy": "credits (voucher on future invoices, 'cannot be reimbursed in cash') \u2014 GPU tier: 10% of monthly amount if 99.0%