Hyperscaler · Seattle, WA
AWS
Hyperscaler selling NVIDIA, AMD and Habana GPU instances on EC2 per second in 39 regions, with On-Demand, Spot, Savings Plans and prepaid Capacity Blocks for ML.
GPU types18with a current offer
Current offers233234 readings on file · last 45 days
H100 from$6.88per GPU-hour, on-demand
Entry price$0.53cheapest GPU-hour listed
Countries63 named regions
Trust Score v0.179/100
- Company & track record73
- Infrastructure control100
- Pricing & billing transparency100
- Reliability signals45
- Security & compliance80
- Customer signal—
Scored Sep 29, 2026 from the provider's own pages. Dimensions not yet measured are excluded from the denominator. Criteria ↓
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%
73 / 100
| Founding year published | not disclosedaws.amazon.com · Sep 29, 2026 | 0 / 4 |
| Headquarters published | Amazon Web Services, Inc.aws.amazon.com · Sep 29, 2026 | 3 / 3 |
| Legal entity named | Amazon Web Services, Inc.aws.amazon.com · Sep 29, 2026 | 4 / 4 |
| Funding or ownership disclosed | Subsidiary of Amazon.com, Inc. (public)aws.amazon.com · Sep 29, 2026 | 4 / 4 |
Infrastructure control
weight 15%
100 / 100
| Capacity model published | operates or rents dedicated capacityaws.amazon.com · Sep 29, 2026 | 10 / 10 |
| Regions published | 3 regions or countriesaws.amazon.com · Sep 29, 2026 | 5 / 5 |
Pricing & billing transparency
weight 20%
100 / 100
| Public prices without an account | yes (read on the pricing page)aws.amazon.com · Sep 29, 2026 | 5 / 5 |
| Billing granularity published | per secondaws.amazon.com · Sep 29, 2026 | 4 / 4 |
| Egress policy published | metered: 100 GB/month free across AWS, then 0.09 USD/GB first 10 TB, 0.085 next 40 TB, 0.07 next 100 TB, 0.05 above 150 TB (to internet from us-east-1); inbound freeaws.amazon.com · Sep 29, 2026 | 4 / 4 |
| Stopped-instance billing published | Compute not billed when stopped; EBS volumes (e.g. gp3 0.08 USD/GB-month) and public IPv4 addresses (0.005 USD/hour, idle or in use) keep billingaws.amazon.com · Sep 29, 2026 | 4 / 4 |
| Public IPv4 policy published | extra: 0.005 USD per public IPv4 address per hour (in use or idle)aws.amazon.com · Sep 29, 2026 | 3 / 3 |
Reliability signals
weight 20%
45 / 100
| Status page | not disclosedaws.amazon.com · Sep 29, 2026 | 0 / 6 |
| SLA published | 99.99% region-level (multi-AZ) and 99.5% instance-level monthly uptime, with service creditsaws.amazon.com · Sep 29, 2026 | 6 / 6 |
| SLA with a numeric uptime commitment | 99.99% region-level (multi-AZ) and 99.5% instance-level monthly uptime, with service creditsaws.amazon.com · Sep 29, 2026 | 3 / 3 |
| Uptime history published | not disclosedaws.amazon.com · Sep 29, 2026 | 0 / 5 |
| Observed availability (30 days) | not measured yet | — |
Security & compliance
weight 15%
80 / 100
| SOC 2 | SOC 1, SOC 2, SOC 3aws.amazon.com · Sep 29, 2026 | 5 / 8 |
| ISO 27001 | ISO 27001 (also 27017, 27018, 27701, 42001)aws.amazon.com · Sep 29, 2026 | 4 / 4 |
| HIPAA or GDPR DPA | Business Associate Addendum availableaws.amazon.com · 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 →
Market report · Oct 25, 2026 · Hyperscalers and general clouds · #1 of the 200 to audit · prices read on the provider's pageWhat the market report notes about AWS
Our internal report lists 248 actors with their zone, segment, status and the prices it could read, dated Oct 25, 2026. Its notes are reproduced as written, in French, with their sources.
- Listed as
- Amazon AWS
- Cross-checkfr
- H100 p5 6,88 $/GPU-h ; B200 14,24 $ ; B300 17,80 $ à la demande ; H200 Capacity Block 4,98 $ après +15 % (04/01/2026) ; région KSA en déc. 2026.
Prices by GPU · 12 GPU types with a current offerWhat AWS 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 | $6.88 | $3.17 | +117% | 3 | on-demandspot | H100 prices |
| A100 80GB | $3.43 | $1.99 | +72% | 1 | on-demand | A100 80GB prices |
| H200 | $7.91 | $4.76 | +66% | 1 | on-demand | H200 prices |
| L40S | $1.86 | $1.26 | +47% | 8 | on-demand | L40S prices |
| B200 | $14.24 | $7.50 | +90% | 1 | on-demand | B200 prices |
| B300 | $17.80 | $8.96 | +99% | 1 | on-demand | B300 prices |
| L4 | $0.80 | $0.80 | +1% | 20 | on-demandspot | L4 prices |
| T4 | $0.53 | $0.51 | +3% | 14 | on-demandspot | T4 prices |
| A100 40GB | $2.74 | $1.98 | +39% | 2 | on-demandspot | A100 40GB prices |
| Tesla V100 32GB | $3.90 | $2.48 | +57% | 2 | on-demandspot | Tesla V100 32GB prices |
| Tesla V100 16GB | $3.06 | $1.71 | +79% | 6 | on-demandspot | Tesla V100 16GB prices |
| A10G | $1.01 | $1.01 | −0% | 16 | on-demandspot | A10G prices |
All offers · 234 readings on fileEvery AWS 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 | |
|---|---|---|---|---|---|---|---|---|
| T4 | 1× T4 | Spot | Australia +16 | Stock unknown | $0.09 | $0.09 | 21:09 UTCfresh | Receipt |
| L4 | 1× L4 | Spot | US | Stock unknown | $0.10 | $0.10 | 21:09 UTCfresh | Receipt |
| L4 | 1× L4 | Spot | US | Stock unknown | $0.13 | $0.13 | 21:09 UTCfresh | Receipt |
| T4 | 1× T4 | Spot | Australia +16 | Stock unknown | $0.13 | $0.13 | 21:09 UTCfresh | Receipt |
| L4 | 4× L4 | Spot | US | Stock unknown | $0.15 | $0.15 | 21:09 UTCfresh | Receipt |
| T4 | 8× T4 | Spot | Australia +16 | Stock unknown | $0.16 | $0.16 | 21:09 UTCfresh | Receipt |
| T4 | 4× T4 | Spot | Australia +16 | Stock unknown | $0.16 | $0.16 | 21:09 UTCfresh | Receipt |
| L4 | 1× L4 | Spot | US | Stock unknown | $0.17 | $0.17 | 21:09 UTCfresh | Receipt |
| T4 | 1× T4 | Spot | Australia +16 | Stock unknown | $0.19 | $0.19 | 21:09 UTCfresh | Receipt |
| L4 | 4× L4 | Spot | US | Stock unknown | $0.20 | $0.20 | 21:09 UTCfresh | Receipt |
| L4 | 1× L4 | Spot | US | Stock unknown | $0.20 | $0.20 | 21:09 UTCfresh | Receipt |
| L4 | 8× L4 | Spot | US | Stock unknown | $0.21 | $0.21 | 21:09 UTCfresh | Receipt |
| A10G | 4× A10G | Spot | UAE +13 | Stock unknown | $0.24 | $0.24 | 21:09 UTCfresh | Receipt |
| L4 | 1× L4 | Spot | US | Stock unknown | $0.26 | $0.26 | 21:09 UTCfresh | Receipt |
| A10G | 1× A10G | Spot | UAE +13 | Stock unknown | $0.28 | $0.28 | 21:09 UTCfresh | Receipt |
| A10G | 4× A10G | Spot | UAE +13 | Stock unknown | $0.29 | $0.29 | 21:09 UTCfresh | Receipt |
| A10G | 8× A10G | Spot | UAE +13 | Stock unknown | $0.30 | $0.30 | 21:09 UTCfresh | Receipt |
| L4 | 1× L4 | Spot | US | Stock unknown | $0.31 | $0.31 | 21:09 UTCfresh | Receipt |
| A10G | 1× A10G | Spot | UAE +13 | Stock unknown | $0.32 | $0.32 | 21:09 UTCfresh | Receipt |
| T4outlier · suspended | 1× T4 | Spot | Australia +16 | Stock unknown | $0.33 | $0.33 | 21:09 UTCfresh | Receipt |
| A10G | 1× A10G | Spot | UAE +13 | Stock unknown | $0.42 | $0.42 | 21:09 UTCfresh | Receipt |
| L4 | 1× L4 | Spot | US | Stock unknown | $0.43 | $0.43 | 21:09 UTCfresh | Receipt |
| A10G | 1× A10G | Spot | UAE +13 | Stock unknown | $0.51 | $0.51 | 21:09 UTCfresh | Receipt |
| T4outlier · suspended | 1× T4 | Spot | Australia +16 | Stock unknown | $0.51 | $0.51 | 21:09 UTCfresh | Receipt |
| T4 | 1× T44 vCPU · 16 GB RAM | On-demand | Germany +3 | Stock unknown | $0.53 | $0.53 | 21:09 UTCfresh | Receipt |
75 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
- 60 seconds per instance launch (Linux, Windows, RHEL, Ubuntu)
- Stopped instance
- Compute not billed when stopped; EBS volumes (e.g. gp3 0.08 USD/GB-month) and public IPv4 addresses (0.005 USD/hour, idle or in use) keep billing
- Egress
- metered: 100 GB/month free across AWS, then 0.09 USD/GB first 10 TB, 0.085 next 40 TB, 0.07 next 100 TB, 0.05 above 150 TB (to internet from us-east-1); inbound free
- Public IPv4
- extra: 0.005 USD per public IPv4 address per hour (in use or idle)
- Billing model
- postpaid invoice
- Minimum deposit
- not published
- Free credits
- Up to 200 USD Free Tier credits for new accounts (100 USD at sign-up, up to 100 USD more), Free plan closes after 6 months
Account
- Payment
- card
- Sign-up
- email, password, card, billing address, phone, additional verification for India-based accounts
- Identity check
- card verification
- Support
- re:Post knowledge center
- Enterprise support
- not published
- SLA
- 99.99% region-level (multi-AZ) and 99.5% instance-level monthly uptime, with service credits
- Status page
- not published
- Published uptime
- not published
Infrastructure
- Regions
3 named regions in 6 countries
us-east-1 (United States) · 39 geographic regions · 124 availability zones (None)- Countries
- United States, Japan, India, United Kingdom, Sweden, Brazil
- Platform
- api, cli, serverless, jupyter, ssh, custom_docker_image, persistent_volumes, snapshots, spot_or_interruptible, bare_metal · SDK : python (boto3), JavaScript, Java, .NET, Go, PHP and others · Terraform : official · Kubernetes : managed · interconnexion : 900 GB/s NVSwitch in P5/P5e/P5en · EFA up to 3200 Gbps · UltraClusters up to 20,000 GPUs · GB200 NVL72 UltraServers · clusters : EC2 UltraClusters, AWS ParallelCluster, EKS, SageMaker HyperPod
- Storage
- volume · 0.08 USD/GB-month (EBS gp3, us-east-1; 3,000 IOPS and 125 MB/s included) · gp2 0.10 USD/GB-monthephemeral · included · Instance-store NVMe on P5/P5en (8 x 3.84 TB) and other families, lost on stop
- Migration out
- EBS snapshots and AMIs are portable within AWS; moving data out costs 0.09 USD/GB after the free 100 GB, though AWS offers free egress to eligible customers leaving AWS entirely (per pricing page).
Security & compliance
- SOC 2
- SOC 1, SOC 2, SOC 3
- ISO 27001
- ISO 27001 (also 27017, 27018, 27701, 42001)
- HIPAA
- Business Associate Addendum available
- GDPR / DPA
- AWS Data Processing Agreement
- Other
- PCI DSS and others
- Legal entity
- Amazon Web Services, Inc.
- Funding
- Subsidiary of Amazon.com, Inc. (public)
GPUBank market study, read on the provider's own pages · 18 pages read · facts marked "not published" were looked for and not foundPages read ↓
From the study · assessmentAWS, in short
What stood out when we read the documents: strengths, weaknesses, who it suits, and what we could not find.
Strengths
- Largest GPU catalogue among the providers reviewed: B300, B200, GB200, H200, H100, A100, L40S, L4, A10G, T4, RTX PRO 6000, RTX PRO 4500, V100 and fractional L4
- Per-second billing with a 60-second minimum, across On-Demand, Spot and Capacity Blocks
- Capacity Blocks let a customer book H100 from 5.191 USD/GPU-hr and B200 from 12.355 USD/GPU-hr for a fixed future window
- 99.99% region-level and 99.5% instance-level SLA with service credits
- SOC 1/2/3, ISO 27001 and HIPAA BAA
- 39 regions; EFA networking and UltraClusters for large training
Weaknesses
- On-demand H100 at 6.88 USD/GPU-hr, H200 at 7.912, B200 at 14.2416 in us-east-1
- Egress 0.09 USD/GB after 100 GB/month; public IPv4 0.005 USD/hour extra; EBS extra
- Large GPUs sold mostly as 8-GPU instances (only p5.4xlarge offers a single H100)
- Postpaid card billing with phone verification; no crypto; PayPal not accepted
- Spot prices for large GPUs can exceed on-demand (p5.48xlarge on the check date)
Ideal for
- Enterprises already on AWS needing GPUs next to their data and compliance scope
- Large training runs using Capacity Blocks or UltraClusters
- Inference fleets on L4, A10G or T4 with managed services around them
Not published
- p2 (K80) and g3s.xlarge prices exist in the feed but their GPU specs are not on the current docs pages read, so they are excluded
- A100 80GB (p4de) form factor not stated
- Savings Plans and Reserved Instance GPU rates not captured
- Founding year and HQ city not stated on pages read
- Spot prices are volatile
- values are a snapshot of the feed on 2026-09-29
- Support plan tiers and prices not captured
- Status page URL not captured
Observations
On-demand pricing page renders instance prices only through a JavaScript region/OS selector; prices here come from the JSON feed that the page loads
Capacity Blocks page lists effective hourly rates per instance and per accelerator by region
Spot page shows the lowest current Spot price per instance type, updated every 5 minutes
Pages read
- aws.amazon.com/ec2/pricing/on-demand/ — Billing model, data transfer prices, 100 GB free egress
- b0.p.awsstatic.com/pricing/2.0/meteredUnitMaps/ec2/USD/current/ec2-ondemand-without-sec-sel/US%20East%20(N.%20Virginia)/Linux/index.json — Hourly on-demand price per instance type
- website.spot.ec2.aws.a2z.com/spot.js — us-east-1 Linux Spot prices
- aws.amazon.com/ec2/spot/pricing/ — Spot mechanics, up to 90% off
- aws.amazon.com/ec2/capacityblocks/pricing/ — Capacity Block rates
- docs.aws.amazon.com/ec2/latest/instancetypes/ac.html — GPU model, count, memory per instance type
- docs.aws.amazon.com/ec2/latest/instancetypes/pg.html — P3, G3 GPU specs
- aws.amazon.com/ec2/instance-types/p5/ — NVSwitch, EFA, UltraClusters, instance store
- aws.amazon.com/ec2/pricing/ — Per-second billing, 60 s minimum
- aws.amazon.com/ebs/pricing/ — gp3/gp2 prices
- aws.amazon.com/vpc/pricing/ — Public IPv4 0.005 USD/hour
- repost.aws/knowledge-center/accepted-payment-methods — Cards, ACH, SEPA, Bacs; PayPal not accepted
- repost.aws/knowledge-center/create-and-activate-aws-account — Payment method, phone verification
- aws.amazon.com/compute/sla/ — 99.99% / 99.5% SLA
- aws.amazon.com/about-aws/global-infrastructure/regions_az/ — 39 regions, 124 AZs
- aws.amazon.com/compliance/programs/ — SOC, ISO, HIPAA BAA, DPA
- aws.amazon.com/free/ — Up to 200 USD credits
- aws.amazon.com/agreement/ — Headquarters
Questions about AWS
What GPUs does AWS offer?
AWS has 18 GPU types with a fresh offer today, for example H100 from $6.88, A100 80GB from $3.43, H200 from $7.91, L40S from $1.86, B200 from $14.24, B300 from $17.80 per GPU-hour on-demand. The full list with prices and market medians is on this page.
How does AWS bill?
Billing granularity: per second. Minimum charge: 60 seconds per instance launch (Linux, Windows, RHEL, Ubuntu) Stopped instance: Compute not billed when stopped; EBS volumes (e.g. gp3 0.08 USD/GB-month) and public IPv4 addresses (0.005 USD/hour, idle or in use) keep billing
How trustworthy is AWS?
Its Trust Score is 79/100 (v0.1), read from its own documents. Strong on infrastructure control, pricing & billing transparency. Weaker on reliability signals. The score measures transparency and published guarantees, not customer satisfaction; every criterion and its source is listed above.
Where are AWS's data centres?
3 named regions in 6 countries: United States, Japan, India, United Kingdom, Sweden, Brazil.
How do you pay AWS, and are there credits?
Payment by card. Free credits: Up to 200 USD Free Tier credits for new accounts (100 USD at sign-up, up to 100 USD more), Free plan closes after 6 months. Identity check: card verification.