Cloud infrastructure · Singapore
Stratos Compute sells cloud infrastructure to AI teams: on-demand GPU and CPU instances, block and object storage, and networking on our public cloud — or the same platform deployed as a private cloud inside your own datacenter.
Running production workloads for
Menlo Research Pte Ltd
Open-source AI research
Jan AI Pte Ltd
Local-first AI assistant
What we sell
Both products are billed in USD and invoiced monthly. Public cloud is self-service and pay-as-you-go; private cloud is a fixed engagement agreed in writing before any work starts.
Rent capacity by the hour from our Singapore-managed fleet. No commitment, no quote, no minimum term.
The same platform, deployed on your hardware or in your colocation, operated by us. Your data never leaves your racks.
The platform
Stratos Compute runs its own OpenStack-based control plane: a self-service console for your engineers, an operator console for pricing and invoicing, and one API behind both.
On-demand instances across NVIDIA RTX PRO 6000, A6000, 4090 and 3090 class accelerators, with GPU quota per project.
Block volumes, snapshots, and S3-compatible object storage on Ceph RGW — including per-key access and static site hosting.
Isolated tenant networks, routers, security groups, floating IPs and managed load balancers.
Kubernetes clusters provisioned and upgraded on your project, plus managed PostgreSQL, MySQL, MariaDB, OpenSearch, Kafka, FerretDB and Valkey.
Every resource is metered hourly and rated against a price plan, with savings plans, promotional credits and account credits.
Automated monthly invoices, card payments through Stripe, bank transfer, and dunning for overdue accounts.
Multi-organisation, multi-project access control with per-project roles and user invitations.
OpenID Connect against Keycloak or your existing identity provider, using authorization-code with PKCE.
A Model Context Protocol endpoint so AI agents can drive the same operations under the same permissions as a human user.
Pricing
List prices in US dollars, billed per hour of actual usage and invoiced monthly in arrears. Prices exclude any tax that applies to you. Private cloud is quoted per engagement.
Per GPU, per hour — added on top of the vCPU and memory of the flavour.
| Accelerator | VRAM | USD / GPU-hour |
|---|---|---|
| NVIDIA RTX PRO 6000 (Blackwell) | 96 GB | $1.99 |
| NVIDIA RTX PRO 4500 (Blackwell) | 32 GB | $0.69 |
| NVIDIA RTX 4090 | 24 GB | $0.65 |
| NVIDIA RTX A6000 | 48 GB | $0.47 |
| NVIDIA RTX 3090 | 24 GB | $0.43 |
| NVIDIA RTX 3080 Ti | 12 GB | $0.29 |
Charged on every instance, GPU or not.
On top of the instances, volumes and load balancers they use.
Database engines: PostgreSQL, MySQL, MariaDB, OpenSearch, Kafka, FerretDB and Valkey. Each engine has its own per-vCPU, per-GB-memory and per-GB-storage rate, shown in the console before you create a cluster.
Usage is metered hourly and rated against your price plan. An invoice is issued at the start of each month for the month just ended.
Card payments are processed by Stripe. Bank transfer is available for invoiced accounts. We never store your card details on our systems.
Savings plans discount long-running capacity, and account credits are applied to an invoice before any card is charged.
Send us the shape of the workload — GPUs, storage, egress, and whether it has to stay on your own hardware. We reply with capacity and a price, usually the same day.
Email yeowhua@stratoscompute.comCompany
Stratos Compute Pte Ltd
UEN 202639575H
Registered office
143 Cecil Street, #06-02GB BuildingSingapore 069542Phone
+65 8084 8126Billing
finance@stratoscompute.com