Private GPU infrastructure,
billed by the second.
Isolated hosts from T4 to B200, provisioned in under a minute. Sign up with an email address, settle in crypto, and keep both your workloads and your identity to yourself.
Tengsuan is built on a single principle: we collect the minimum required to run and bill your workloads, and nothing more. No identity verification to rent a GPU. No card on file. No inspection of what you run. Compute is provisioned when you need it and destroyed when you are done.
Designed to retain as little as possible.
Every component starts from one question: what is the least we need to know to run your workload well?
Ephemeral compute
Each sandbox or run is an isolated container on a dedicated accelerator. When it stops, its filesystem is destroyed. Persistent data belongs on a volume you control; nothing else is retained.
Crypto-settled billing
Verification is a deposit, not a document. Usage is metered per second and invoiced monthly, payable in the coin and network of your choice to a unique address.
Encrypted secrets
API keys and tokens are encrypted at rest and injected only into containers you launch.
Provisioned in seconds
Hosts are ready in well under a minute. Provisioning time is not billed.
What we retain, and what we do not.
The complete list, in plain language and consistent with our privacy policy.
Email address and password hash
This is the entire identity record. No phone number, government ID or card is collected, and any email address may be used.
Usage ledger
GPU type, seconds, cost, and the deposit address that settled it. Required for accurate billing and nothing further.
Security events, for twelve months
Sign-in events and IP addresses are retained to detect account takeover, then deleted.
Contents of your hosts
We do not read, scan or retain what runs inside your containers. Run logs expire automatically after seven days.
Payment identity
Crypto deposits carry no name. We never receive card details, and usage data is never sold or shared.
Data after account closure
Closing the account deletes its records. Only invoice amounts and payment references are kept, for as long as financial regulation requires, with nothing else attached.
From sign-up to first workload in four steps.
Hover or tap each step for detail.
Create an account
An email address and a password. There are no verification calls or questionnaires.
Make a deposit
Send USDT, USDC, BTC, ETH or SOL to a one-time address. A first deposit of $25 or more verifies the account and is credited in full.
Run
Choose a GPU from T4 to B200, an image, a volume and any secrets. Connect over SSH or submit a script from the CLI. Usage is metered per second.
Terminate
Stop the host and its filesystem is destroyed. Return whenever you need compute again.
GPUs available on demand.
Rates include the default vCPU and memory shape. Capacity is provisioned on demand and billed only while a host is running.
NVIDIA T4
Turing architecture, 16 GB of VRAM, up to 8 per host. 4 vCPU and 16 GB of RAM in the default shape.
NVIDIA A100 80GB
Ampere architecture, 80 GB of VRAM, up to 8 per host. 8 vCPU and 48 GB of RAM in the default shape.
NVIDIA H100 80GB
Hopper architecture, 80 GB of VRAM, up to 8 per host. 12 vCPU and 64 GB of RAM in the default shape.
NVIDIA H200 141GB
Hopper architecture, 141 GB of VRAM, up to 8 per host. 12 vCPU and 96 GB of RAM in the default shape.
NVIDIA B200 180GB
Blackwell architecture, 180 GB of VRAM, up to 8 per host. 16 vCPU and 128 GB of RAM in the default shape.
Private compute,
on your terms.
Create an account, make a deposit, and launch your first host within minutes.
Get started