GPUBank vs Thunder Compute
Thunder Compute is a US developer GPU cloud that runs its own virtualization layer: instances use GPUs attached over the network from a shared pool. Its H100 SXM costs $3.20 per GPU-hour on demand; GPUBank charges $1.85 on a GPU attached to the VM. Thunder is cheap on older cards, with the RTX A6000 at $0.35 and the A100 80GB at $1.09, but it bills per minute until deletion and runs in North America only.
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
Thunder Compute
San Francisco, United States
H100 SXM
$3.20 / h
- RTX A6000 at $0.35 and L40 at $0.79 per GPU-hour
- A100 80GB at $1.09 per GPU-hour on 1 or 2 GPUs
- Per-second H100 sandboxes with a TTL for agent jobs
Thunder Compute
Developer cloud
United States27 points compared13 of their pages readprices last checked on 29 September 2026page updated 2 November 2026
The verdict
Low A6000 and A100 prices at Thunder Compute. Cheaper H100, stop billing and more regions at GPUBank.
Thunder Compute suits developers who prototype on an RTX A6000 or A100 from VS Code or an AI agent, and do not need the newest GPUs. Its GPUs are virtualized over the network, and its docs list features that do not work, such as managed memory, CUDA MPS and GPU profilers. Instances bill per minute from creation to deletion, with no stop state; GPUBank bills per second and keeps only the disk on the bill when you stop. On the H100 SXM, GPUBank costs 42 % less, runs in four regions and takes crypto.
Thunder ComputeGPUBank 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 | Thunder Compute per hour | GPUBank per hour | Difference | 100 h at Thunder Compute | 100 h at GPUBank |
|---|---|---|---|---|---|
 H100 SXM80 GB |
$3.20 | $1.85 | −42 % | $320.00 | $185.00 |
A fine-tune, 40 h on an H100 SXM
Thunder Compute
$128.00
GPUBank
$74.00
Saves $54.00 at GPUBank (−42 %)
02Their catalogue
Every GPU Thunder Compute lists.
4 GPU models listed by Thunder Compute, each at its lowest public price per GPU-hour. 1 of them is also rented at GPUBank (in blue).
- H100 SXM80 GB · on demand$3.20
- A100 80GB80 GB · on demand$1.09
L4048 GB · on demand$0.79
RTX A600048 GB · on demand$0.35
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.
Thunder Compute
2 h 14 m 03 s on H100 SXM
- GPU · $3.20 / h
- $7.15
- vCPU and RAM
- Included
- Storage
- $0.22 per GB-month above 100 GB per GPU
- Egress
- Included
- Public IPv4
- Dynamic IP address
- Billing granularity
- Per minute; sandboxes per second
- 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.
Thunder Compute
No stop state, billed until deleted.
To pause, snapshot at $0.05 per GB-month and delete.
GPUBank
Only the persistent disk, $0.03 per GB-month. Compute, IPv4 and scratch cost nothing.
04Where the GPUs run
1 country against 4, and where they are.
Countries where Thunder Compute 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.
United States
Paris
Frankfurt
Ashburn
Singapore
Thunder Compute1
- United States
GPUBank4
Paris
Frankfurt- Ashburn
Singapore
Thunder Compute · 1 countryGPUBank · 4 regions
05Paying and signing up
How each one takes your money.
Thunder Compute
Payments and sign-up
Payment methods
Cards
No cryptoMinimum top-up Not published
To start
- 1Email
- 2Card or preloaded credit
Billing modelCard auto-pay or prepaid credit
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.
| Thunder Compute | GPUBank | |
|---|---|---|
| SOC 2 | Not published | Not published |
| ISO 27001 | Not published | Not published |
| HIPAA | Not published | No |
| GDPR DPA | DPA available | Published DPA, applies automatically |
| Availability target | None; availability not guaranteed | 99.9 % monthly, per region |
| Status page | status.thundercompute.com | Public status page |
| Support | Discord and email | 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 Thunder Compute pricing page captured on 29 September 2026, and GPUBank's live board. Cookie banners and chat bubbles are closed; nothing else is edited.
- 1Per GPU-hour prices are public, with 1x, 2x, 4x and 8x tabs and a cost calculator.
- 2The page compares its A100 price with Runpod, CoreWeave, Lambda, AWS and others.
- 3The calculator adds extra vCPU at 0.04 USD per vCPU-hour and extra disk by the GB-hour.
- 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
27 points, read off both sites.
GPUBank aheadThunder Compute aheadEven
| Criterion | GPUBank | Thunder Compute |
|---|---|---|
| Price and catalogue | ||
| H100 SXM, per GPU-hourH100 sandboxes cost $2.95 per GPU-hour for the GPU alone, with vCPU, RAM and disk on top. | $1.85, all-in | $3.20 on demand |
| Cheapest GPU-hour | $0.23, RTX 4090 | $0.35, RTX A6000 |
| GPU models listed | 10 | 4 |
| What the GPU price includesExtra vCPUs above the base are charged. | vCPU, RAM, NVMe, IPv4, traffic | Base vCPU and RAM, 100 GB disk per GPU |
| Billing | ||
| Billing granularity | Per second | Per minute; sandboxes per second |
| Minimum charge | None | Not published |
| Stopped instanceTo pause, snapshot at $0.05 per GB-month and delete. | Disk only | No stop state, billed until deleted |
| Persistent storage | $0.03 per GB-month | $0.22 per GB-month above 100 GB per GPU |
| Egress traffic | Included | Included |
| Public IPv4 | Included | Dynamic IP address |
| Billing model | Prepaid balance | Card auto-pay or prepaid creditAhead |
| Payment and sign-up | ||
| Crypto payments | Bitcoin, USDT, Ethereum and 12+ more | No |
| Card payments | No | Credit cardsAhead |
| Minimum deposit | $25 | Not published |
| Platform | ||
| API and CLI | REST API, CLI, Python SDK | REST API, tnr CLI, VS Code extension, MCP |
| Terraform | Official provider | No |
| Serverless GPUs | No | H100 sandboxes, billed per secondAhead |
| Managed Kubernetes | No | No |
| Ready-made templates | 25 | Ollama, ComfyUI and others |
| Interruptible or spot | Yes, 40 % off | No |
| GPUs per instance | Up to 8 | Up to 8; reserved clusters of 8 to 512+Ahead |
| Multi-node interconnectDocs list managed memory, CUDA MPS, FFmpeg GPU acceleration and profilers as unsupported. | Not offered | Network-attached virtualized GPUs |
| Network and regions | ||
| Regions | 4 | North America only |
| Countries | France, Germany, US, Singapore | United States |
| EU regions | Paris, Frankfurt | None |
| Trust and support | ||
| Availability target | 99.9 % monthly | None; availability not guaranteed |
| Compliance reports | None published | No SOC 2 or ISO; DPA published |
| Public status page | Yes | status.thundercompute.com |
| Support | Tickets from a $25 balance | Discord and email |
09The right fit
Different needs. A clear choice.
What Thunder Compute does better
- RTX A6000 at $0.35 and L40 at $0.79 per GPU-hour
- A100 80GB at $1.09 per GPU-hour on 1 or 2 GPUs
- Per-second H100 sandboxes with a TTL for agent jobs
- VS Code extension, tnr CLI and an MCP server
- No egress fees and 100 GB of disk per GPU included
Choose GPUBank if
- H100 SXM at $1.85 per GPU-hour, against $3.20
- GPUs attached to the VM, not over the network
- Per-second billing, and a stopped instance costs only its disk
- Paris, Frankfurt, Ashburn and Singapore; Thunder is North America only
- B200, H200, MI300X, RTX 5090 and RTX 4090, which Thunder does not list
Moving from Thunder Compute takes an afternoon.
Thunder charges no egress fee, and its instances run standard Linux with Docker, so workloads move to any VM. Snapshots stay on Thunder, so copy data out before you delete.
- 1Launch the same GPU on GPUBank with a persistent disk in the region you need.
- 2Copy your data from the Thunder instance with rsync or scp over SSH.
- 3Run the same Docker image or pick the matching template, then delete the Thunder instance and its snapshots.
# 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.
Yes. Thunder Compute lists the H100 SXM at $3.20 per GPU-hour on demand, and $2.95 in sandboxes before vCPU, RAM and disk. GPUBank charges $1.85 per GPU-hour with vCPU, RAM, NVMe, IPv4 and traffic included.
They are physical GPUs, attached to your instance over the network through Thunder's virtualization layer. Its docs list incompatibilities: managed memory, CUDA MPS, FFmpeg GPU acceleration and GPU profilers. GPUBank GPUs are attached to the VM.
No. Instances bill per minute from creation until deletion. To pause, you snapshot the disk at $0.05 per GB-month and delete the instance. On GPUBank a stopped instance costs only its persistent disk, at $0.03 per GB-month.
It depends on the size. Thunder lists the A100 80GB at $1.09 per GPU-hour on 1 or 2 GPUs and $1.49 on 4 or 8, with the form factor not published. GPUBank charges $0.95 for an A100 SXM 80GB at any size up to 8 GPUs.
No. It takes card auto-pay or preloaded credit, and the credit is non-refundable. GPUBank is paid from a USD balance topped up in Bitcoin, USDT, Ethereum and more than a dozen other coins.
Without. Every price on this page excludes VAT and sales tax, for Thunder Compute 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 Thunder Compute was read on its own pages or official documentation. Prices last checked on 29 September 2026.
- 01thundercompute.com/pricingInstance, sandbox, storage, reserved prices, FAQ (per-minute billing, egress)
- 02thundercompute.com/docs/billing.mdAuto-pay card, preload credit, threshold charges, non-refundable credit
- 03thundercompute.com/docs/restrictions.mdCountry restrictions, one account per user, crypto mining ban
- 04thundercompute.com/docs/technical-specs.mdLocation, disk, network, dynamic IP
- 05thundercompute.com/docs/guides/stopping-instances.mdNo native stop
- 06thundercompute.com/docs/compatibility.mdKnown incompatibilities
- 07thundercompute.com/docs/faq/faq.mdNetwork-attached GPU pool, EC2 statement
- 08thundercompute.com/terms-and-conditionsLegal entity, Delaware, payment, no uptime guarantee
- 09thundercompute.com/sub-processorsAzure, Stripe
- 10thundercompute.com/products/instancesSnapshots, restore time
- 11thundercompute.com/Series A funding
- 12thundercompute.com/missionResearch start 2022, GPU virtualization
- 13status.thundercompute.comUptime
What we could not verify (6)
- A100 80GB form factor not published
- Reserved prices and term length (contact sales)
- Exact data center locations (docs: North America; sub-processors: Azure primarily Western US; FAQ: 'under the hood they are EC2 instances')
- Preloaded credit payment rails (only card/Mastercard documented); no crypto
- Base vCPU included per GPU and extra vCPU price only implied by calculator
- Founding year and HQ city not stated
In short
GPUBank vs Thunder Compute, 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
- Thunder Compute: $3.20 on demand
GPUBank: $1.85, all-in - Cheapest GPU-hour
- Thunder Compute: $0.35, RTX A6000
GPUBank: $0.23, RTX 4090 - Billing granularity
- Thunder Compute: Per minute; sandboxes per second
GPUBank: Per second - Stopped instance
- Thunder Compute: No stop state, billed until deleted
GPUBank: Disk only - Persistent storage
- Thunder Compute: $0.22 per GB-month above 100 GB per GPU
GPUBank: $0.03 per GB-month - Egress traffic
- Thunder Compute: Included
GPUBank: Included - Crypto payments
- Thunder Compute: No
GPUBank: Bitcoin, USDT, Ethereum and 12+ more - Card payments
- Thunder Compute: Credit cards
GPUBank: No - Minimum deposit
- Thunder Compute: Not published
GPUBank: $25 - GPU models listed
- Thunder Compute: 4
GPUBank: 10 - Regions
- Thunder Compute: North America only
GPUBank: 4 - Serverless GPUs
- Thunder Compute: H100 sandboxes, billed per second
GPUBank: No - Availability target
- Thunder Compute: None; availability not guaranteed
GPUBank: 99.9 % monthly - Compliance reports
- Thunder Compute: No SOC 2 or ISO; DPA published
GPUBank: None published - Support
- Thunder Compute: Discord and email
GPUBank: Tickets from a $25 balance - Headquarters
- Thunder Compute: San Francisco, United States
GPUBank: Tallinn, Estonia
How this page was made
Every price and policy of Thunder Compute 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. Thunder Compute, its logo and its website belong to their owner and are shown only to identify the service compared. GPUBank is not affiliated with Thunder Compute.
