Neysa has a lot going for it. The AI cloud provider turned unicorn earlier this year after raising $1.2 billion led by Blackstone. Its founder Sharad Sanghi was recognised as AIM Person of the Year for building the infrastructure to power India’s AI ambitions.
Now, it’s building atop its full-stack AI cloud platform.
Neysa has partnered with US-based telecom technology company Mavenir to combine GPU infrastructure with an AI software layer that enterprises and service providers can use to deploy and manage AI workloads in India.
The partnership brings together Neysa’s GPU infrastructure and AI cloud platform with Mavenir’s AI orchestration, agent workflows, security, policy controls, and token-level metering. The companies are positioning the combined stack as a way for enterprises to move AI projects from experimentation into production without having to assemble the infrastructure and software layers separately.
“The next phase that we are seeing is about getting AI into production, getting control over performance, security, governance, and obviously, cost at scale,” Xavier Kurian, Chief Revenue Officer at Neysa, tells AIM.
From GPU Capacity to AI Platform
Neysa plans to deploy 20,000 GPUs across India. Kurian says a portion of that capacity will be consumed directly as GPUs, but many enterprises and service providers need a broader platform to use the infrastructure effectively.
“We’re very efficient at going and building out the capacity at scale, making it usable, observable, cost-efficient,” Kurian says. “This partnership is about putting that layer on top, so that enterprises and service providers can get that execution layer for AI.”
Bejoy Pankajakshan, Chief Technology and Strategy Officer at Mavenir, says enterprises lack the software platform needed to use GPUs efficiently.
“What we’ve developed is an AI software stack that allows you to securely access GPU-based training and inferencing,” Pankajakshan says. “It allows you to set the right policies and then monetise using token counting.”
The companies will also use Neysa’s infrastructure internally. Mavenir previously installed its own GPUs across R&D locations for product development. Under the partnership, Neysa will provide the infrastructure, power, and GPUs, while Mavenir develops its AI products on top of the platform.
Orchestration and Token Metering
Mavenir’s AI orchestration layer will manage how inference requests are routed across available infrastructure.
For example, an AI request from a banking application could be routed based on factors such as where the user’s previous conversation context is stored, available capacity, latency, and cost. The orchestrator does not sit directly in the inference path. Instead, it guides the AI gateway and platform.
“The AI orchestrator is not going to be in the path of the actual inferencing,” Pankajakshan says. “It’s going to guide the AI gateways and the AI platform for the path to take. So you get the fastest response, or your cost to be the lowest.”
The platform will also support tenant onboarding and provision infrastructure for operators and service providers.
Token metering is another key component. As training and inference workloads run, the platform can track token consumption and support token-based charging, similar to how telecom operators have historically charged for voice minutes or data usage.
The focus on token metering comes as enterprises become more conscious about tracking their AI costs and avoiding tokenmaxxing. According to a recent OpenRouter analysis of over 380 trillion tokens, agentic workflows and tool-calling requests now drive over 52% of token consumption, causing significant cost volatility for enterprises.
Mavenir says the tokenomics layer can interface with existing billing systems, meaning enterprises and operators would not need to completely rebuild their billing infrastructure to introduce token-based charging.
Mavenir’s AI Products
The partnership will initially support Mavenir’s existing AI products, rather than creating an entirely new product portfolio. These include AI-based service assurance tools for telecom operators, which can help troubleshoot networks in real time, identify root causes, and correlate network events.
Mavenir is also working on voice AI products, including translation services already being offered to operators. Security is another focus, with the company developing solutions around emerging threats such as deepfakes and voice cloning.
Pankajakshan says the companies are also looking at a broader evolution of Mavenir’s service assurance products.
“The one that we have launched right now is the agentic service assurance product,” he says. “That would evolve into a full-blown intent orchestrator that can do end-to-end management of full networks.”
Mavenir expects to develop and launch additional products on the infrastructure over the next six to eight months.
Sovereignty is Key
The companies are also positioning the partnership around sovereign AI, particularly for Indian enterprises and AI developers that need greater control over where data and infrastructure reside.
Pankajakshan says the entire software control plane for this deployment will run within India, rather than relying on a control layer connected to infrastructure outside the country.
“It’s not like a hyperscaler-provided sovereign setup where you could have the control part of it still connected to the US,” he says.
For US-based Mavenir, infrastructure-level sovereignty is essential for its more than 3,000-strong R&D workforce in India.
Kurian says sovereignty ultimately comes down to where data resides, how it moves through the system, who can access it, and which laws and jurisdictions apply.
Beyond Telecom
Although Mavenir has deep roots in telecom, it wants to serve a broader enterprise market. Mavenir expects operators to be an important route to market, including for smaller businesses, while the underlying platform can also serve enterprises and other cloud providers.
“This is something that will make AI usable for a huge segment of enterprises,” Pankajakshan says. “The operators will go take that down to the SMBs and the rest of the market.”
Kurian describes the partnership as complementary to Neysa’s infrastructure strategy, with Neysa providing the GPU layer, cloud environment, and sovereignty, while Mavenir adds the software platform needed to operationalise AI workloads.
Your reaction