NinjaTech AI Inc. launched Ninja Enterprise, letting large companies run AI employees inside their own cloud environments for a fixed yearly fee that includes GPUs. The offer targets corporate projects that stall after the pilot stage because token-based pricing is hard to forecast. Long-running agents worsen this, since charges keep rising for every hour of autonomous work. A predictable annual bill is meant to let AI spread across the organization once testing ends.
Fixed fee covers software, cloud and GPUs
For customers running the platform in their own cloud, the fixed price covers the computing power behind the software, which NinjaTech has not supplied before. Buyers need neither a separate GPU contract nor servers to buy. NinjaTech sources GPU and inference capacity through Microsoft working with Fireworks AI or through Amazon Web Services, reserving it on a single-tenant basis. Packages of 100, 500 or 1,000 AI employees can start with a pilot before an annual capacity contract, with more capacity on demand.
The agents take a goal and work through it unattended, with staff contact through Slack and Microsoft Teams. Workloads run on open-weight models on the reserved capacity, which NinjaTech says costs about a tenth as much as comparable deployments on frontier models from the big AI labs. Customers can still move any workflow to a model from Anthropic or OpenAI when that fits the task better. For work that cannot touch an external network, a second version runs air-gapped on customer-supplied hardware, and in both setups data is never pooled or used to train an outside model.
The move addresses pilots that work technically but stall when usage scales and finance lacks a yearly number. Most vendors keep AI on a usage meter, Chief Executive Babak Pahlavan said. Implementation is included in the same contract, delivered for healthcare customers by Optimum Healthcare IT and for others by Infosys. Infosys invested through the Infosys Innovation Fund in August and agreed to act as exclusive professional services partner for at least a year. NinjaTech was co-founded in 2022 by Pahlavan after 11 years at Google, with SRI Ventures and DCVC as early backers.
What this changes for enterprise AI budgets
For companies adopting AI, the change is mainly in budgeting and procurement. Finance plans against annual capacity instead of a token meter that climbs with agent hours, while IT avoids lining up a separate compute vendor. Single-tenant reservation keeps workloads isolated in the company own cloud without server purchases, and implementation arrives under one contract. The entry point of 100 AI employees suits organizations able to spread agents across teams. Smaller firms may find that starting size above near-term needs, making utilization and internal support the deciding factors.
The fixed fee still leaves questions for selection. The tenth-cost claim for open-weight models is a vendor figure that depends on workload mix, especially if work shifts to Anthropic or OpenAI models for quality reasons. The air-gapped option requires customer-supplied hardware, which alters total cost. Data separation promises need contract terms on retention, access and audit. Before committing, buyers should clarify what the pilot covers, how extra AI employees are priced, and how unused or insufficient reserved capacity is handled.
The test is whether fixed-scope pilots convert to annual capacity contracts and whether workloads stay mainly on open-weight models. Expansion to 500 or 1,000 AI employees without heavy switching to frontier models would support the bundled economics. Frequent moves to Anthropic or OpenAI, or preference for air-gapped installs on own hardware, would show where the fixed-fee approach meets its limits.
