E2E Networks launches Blaze AI inference service with Nvidia

Launching Blaze AI in India

E2E Networks recently introduced Blaze, a production-ready inference service designed for open AI models. Inference is the process where an AI model takes in new data and generates an output, such as a response to a prompt. This service allows developers and enterprises to access leading open models through a single, OpenAI-compatible endpoint.

The launch arrives at a time when many Indian AI companies are moving beyond the pilot phase. As these companies begin to deploy AI agents, the cost and speed of inference become critical factors for the viability of their products. Blaze aims to solve these problems by offering a pay-per-token model, meaning customers only pay for the specific amount of data processed.

Tarun Dua, the founder and managing director of E2E Networks, stated that Blaze is where customers can run their production inference. The service is specifically built to handle inference-heavy applications. These include voice services, coding agents, document AI, image and video generation, and various language applications.

Nvidia-powered infrastructure

Blaze is built on top of Nvidia GPU infrastructure and utilizes open-source components from the Nvidia DSX OS. This software layer is part of the Nvidia DSX AI Factory Platform. By using these components, E2E Networks can provide a robust environment for running frontier-class open-weight models on a Nvidia Cloud Partner (NCP) platform.

Several specific Nvidia technologies work together to manage the workload. Nvidia Dynamo handles distributed model serving, which involves spreading the model across multiple units. It manages things like KV cache-aware routing and disaggregated prefill and decode. It also ensures fast data transfer between GPUs using Nvidia NIXL.

Additional coordination comes from Nvidia Grove and KAI Scheduler, which manage inference workloads across the E2E Kubernetes fleet. To ensure security, Nvidia BlueField DPUs, Nvidia DOCA, and the Nvidia Infra Controller enforce isolation at the hardware level. This ensures that different customers' data remains completely separated from one another.

Sovereign compute and security

A major focus of the Blaze service is providing sovereign compute for the Indian market. Sovereign compute refers to computing resources that are hosted within a specific country's borders and governed by its local laws. This is a significant requirement for many organizations that cannot move sensitive data to external clouds.

Because of this focus on sovereignty, the service is particularly relevant for public-sector bodies, healthcare providers, and banks. These industries often have strict data rules that prevent them from using general-purpose cloud providers. The hardware-level isolation mentioned earlier ensures that these entities can use high-performance AI while remaining compliant with privacy regulations.

Vishal Dhupar, Managing Director for South Asia at Nvidia, noted that access to AI computing and inference services is essential to advancing India’s AI ecosystem. By providing these tools locally, E2E Networks and Nvidia are helping to build a domestic infrastructure that supports the growth of Indian technology companies.

Expanding AI factory capacity

To support the Blaze service, E2E Networks is significantly expanding its AI factory capacity in India. The company is growing its capacity to approximately 14.6 MW. This expansion is designed to meet the increasing demand for high-performance computing power required to run large-scale AI models.

The infrastructure includes a 1,024-GPU Nvidia Blackwell GPU cluster that has been live since May 2026. Blackwell is a series of high-performance chips designed by Nvidia for large-scale AI training and inference. The company is also adding 5.4 MW of additional Blackwell capacity to its portfolio.

This new capacity includes HGX B300 systems and other Blackwell GPUs. By scaling the physical hardware, E2E Networks can host more customers on the Blaze platform simultaneously. This ensures that as more Indian companies move into production, the underlying infrastructure can handle the increased demand for tokens and low-latency responses.