Beyond hyperscalers: Why firms are building alternative AI stacks

Technology companies are building alternatives to global hyperscalers by combining domestic computing infrastructure with AI software, orchestration and governance tools. The aim is to give enterprises greater control over where their AI workloads run, how data is handled and how models are deployed.

Two partnerships announced in September highlight this approach. IBM and Yotta Data Services today announced the general availability of a sovereign agentic AI platform for Indian organisations. A week earlier, Mavenir and Neysa partnered to offer AI infrastructure and orchestration for telecom operators, enterprises and neocloud providers.

The partnerships come as demand for AI-ready infrastructure grows in India. Gartner estimates that public cloud spending in the country will rise 28.1% to $17.5 billion in 2026, from $13.7 billion in 2025. Spending on infrastructure-as-a-service (IaaS), which provides computing and storage resources, is expected to grow 40% to $6.26 billion. The figures point to rising demand for cloud infrastructure, though they do not necessarily indicate a shift away from global cloud providers. The case for alternative AI stacks is likely to depend on the workload, with factors such as latency, cost, data control and regulatory requirements influencing deployment decisions.

IBM, Yotta bring agentic AI to sovereign cloud

The IBM-Yotta platform combines IBM’s watsonx Orchestrate, which acts as the control plane for managing and governing AI agents, with Yotta’s Shakti Cloud, which provides computing capacity, GPUs, networking and security.The platform is designed to let organisations deploy and manage AI agents while keeping data, operations and governance controls within India. It is available through Yotta’s cloud regions in Panvel and Greater Noida, and can be used for workflows such as security operations, document processing and HR automation.

“With regulations evolving in India, digital sovereignty and open-source models are becoming a defining requirement for organisations as they scale AI from pilots to production. The priority is no longer just what AI can do, but how securely, transparently and compliantly it can be deployed,” said Sandip Patel, managing director, IBM India and South Asia, in the press note.

Mavenir, Neysa combine compute and AI software

Mavenir and Neysa’s partnership takes a different approach. Neysa provides GPU infrastructure and the AI cloud environment, while Mavenir brings AI orchestration, agent workflows, security and policy controls, along with token-level metering and billing. Xavier Kurian, chief revenue officer at Neysa, said the partnership brings together capabilities that the two companies have developed separately. “Neysa provides the infrastructure and the knowledge on clusters and scalability and performance and a whole bunch of those things. Mavenir provides the pieces that they are strong with,” Kurian said.

The partnership is intended to help companies move beyond experimentation and deploy AI in production. “At the end of the day, the goal is to drive adoption in the market. And that's not going to happen without the entire stack,” he said. 

Neysa already has its own AI platform, Velocis, which manages infrastructure, networks, performance tuning and guardrails. Kurian said the Mavenir partnership is meant to complement that offering rather than replace it. “This is not A or B. It is A and B,” he said, explaining that Mavenir brings experience in telecom-grade applications and software that Neysa can integrate with its infrastructure.

The companies also expect to address the cost of running AI models. Mavenir's orchestration tools can route queries to different models depending on the task, potentially reducing the need to use more expensive frontier models for every request. However, Kurian said it was difficult to quantify the savings because costs depend on factors such as model selection and quantisation. “The methods on how to do it are what this relationship is about,” he said.

Neysa has announced plans to deploy 20,000 GPUs across India, Kurian said, adding that capacity expansion would be driven by demand. The commercial model for the partnership, including how customers would procure the services, is still being worked out.

Where alternative AI infrastructure could gain ground

The shift towards alternative AI stacks is likely to be selective, according to Hari Balaji, partner, technology consulting at EY India. “Expect the alternatives to gain ground in specific workloads where a good reason exists for the switch,” Balaji said. Telecom operators are an early market because they already have distributed infrastructure and a reason to offer AI services on it. Industrial sites and factory floors are another potential use case, particularly when sending large volumes of video or Internet of Things data to a central cloud creates latency.

Balaji also pointed to real-time news analysis for trading signals and routine, high-volume AI tasks, such as updating large numbers of product listings on e-commerce and fast-fashion platforms. In these cases, economics, as well as latency, could encourage companies to consider alternatives. Beyond specific workloads, Balaji expects original equipment manufacturers to offer enterprises sovereign AI deployment options, with data, governance and infrastructure domiciled and controlled in India.

Kurian said the partnership is intended to offer an alternative to the existing AI infrastructure ecosystem, dominated by large cloud providers. “The golden goal is to give an alternative to the current ecosystem that exists,” he said in the interview. He added that the market is likely to remain hybrid, with some companies opting for GPU infrastructure alone and others seeking a more integrated stack.