NinjaTech launches SuperNinja Enterprise with dedicated GPUs

NinjaTech AI has launched SuperNinja Enterprise, bringing autonomous AI employees, dedicated GPU capacity and software deployed in a customer’s cloud account into one fixed-term contract. NinjaTech says customers pay one rate per AI employee for the contract term instead of per-token charges for open-weight model use. Access to frontier models is available separately through its gateway.

According to NinjaTech, the included GPU capacity is single-tenant, reserved in the United States for the customer’s contract term and supplied as capacity rather than hardware the customer buys. The company says the stack can run behind the customer’s firewall in an Azure, Amazon Web Services, Google Cloud or Oracle account. It also says the GPU spending may count toward existing Azure MACC, AWS EDP or Google Cloud CUD commitments.

NinjaTech describes open-weight model use as including unlimited tokens without rate limits or overage fees. It claims its open-weight inference stack costs about one-tenth as much as frontier models for the same work. That comparison has not been independently validated with cost analysis, benchmarks or customer billing data.

For administration and security, NinjaTech says the stack supports SSO and SCIM, role-based access control, audit trails, and encryption in transit and at rest. The company also claims customer files, memory, connectors and keys remain in the customer’s tenant, with model calls as the only outbound traffic. Those deployment and security assertions have not been independently tested.

NinjaTech says high-impact agent actions require human confirmation and that people must verify output in regulated uses. The company describes deployments as HIPAA-eligible and says its SOC 2 Type II work remains in progress. Its official page does not publish a dollar price, minimum AI-employee count, contract length, GPU model, GPU count or reserved-capacity amount.