Tata Comm’s AI cloud aims to compete on cost with global hyperscalers

AS Lakshminarayanan, CEO and MD, Tata Communications

Tata Communications is betting big on its AI cloud offering Vayu, which looks to compete with global hyperscalers through cost efficiencies amid macroeconomic uncertainties plaguing the tech budgets of customers.

Vayu consolidates the company’s portfolio of cloud and AI offerings under one brand, bringing together its platform-as-a-service (PaaS) and infrastructure-as-a-service (IaaS) portfolio which includes an AI studio, API management, GPU-as-a-service, etc.

AS Lakshminarayanan, the CEO and MD of the telecom-focussed company, said that Tata Communications wants to become a `communications tech' firm that caters to customers across sectors and industries. In that regard, the company already has an ongoing collaboration with Tata Consultancy Services (TCS) where its infrastructure is deployed in the latter’s projects.

Tata Communications also has a partnership with NVIDIA for its GPUs. It is one of the major Indian companies aiming to offer AI cloud computing services to enterprises using high-end NVIDIA Hopper and Blackwell chips. In October 2024, it unveiled its AI infrastructure with NVIDIA's accelerated computing.

As the company looks to compete with global cloud service providers such as AWS, Microsoft Azure, and Google Cloud, it claims its biggest differentiation is cost.

“Compared with large cloud service providers, Vayu reduces costs by 15-25 percent, with no data egress charges (for transferring data to the internet or another cloud provider) or other hidden fees,” the company said in a statement.

“If you look at the private cloud players, barring the hyperscalers, whether it's CntrlS or Sify, we are the most prominent  in this space. We have managed the cloud for the last 10 years. We have the pedigree, can build our own software, and we have our own network,” said Lakshminarayanan.

“We bring the advantage of our multi-cloud connect. Today, data might be sitting in Azure, AWS, on-prem (on the client's premises), and in our AI cloud. We'll be able to unify all the data and use it for AI training and inference. We have the inferencing engine at the edge. We are running 20 edge nodes worldwide, which our media business is using,” he added.

Explaining the cost benefits, Bharat Gorti, Executive Vice President, Cloud and Cybersecurity Services, Tata Communications, said that in enterprises, at least six to 10 people are assigned to look at AI and cloud computing costs, and that continues to rise.

"With our built-in FinOps (financial operations) automation, financial efficiency is baked into your cloud as you scale. That’s where we want to optimise, and we have success stories where we have achieved about 30 percent improvement in price performance,” Gorti said.

“We have data centres in the UK,  US, and Singapore, which offer a cost-effective environment,  are purpose-built, and workload specific,” he added.

Vayu comprises industry-specific clouds that align with compliance and certification needs, and lets  customers use a public and / or private cloud. For instance, it has a dedicated government cloud for public sector entities, and a financial services cloud that's in tune with RBI and SEBI regulations, the DPDP (Digital Personal Data Protection) Act, etc.

Building SLMs, AI use cases

Gorti shared that the company offers GPU resources on rent depending on the customer's computation needs, and an AI model studio to experiment with use cases, both of which help enterprises cut costs.

“LLM builders need a lot of GPUs, but enterprises actually need between 2 to 12 GPUs to build their use cases,” he said, adding that their target customer base includes startups, academia, and MSMEs.

In GPU services, the company’s rivals include Yotta Infrastructure and enterprise AI startup Neysa. Tata Communications was among the 10 companies selected for the final bidding process to procure 10,000 GPUs under the IndiaAI Mission.

Lakshminarayanan said that the company is also helping startups build large language models (LLMs) and small language models (SLMs) through its Kaleyra AI offering.

“We are working on some SLMs in a partner-led model, which we are integrating with what we are doing. For our own internal use, we have an AI COE (centre of excellence). We have launched chatbot-like features and capabilities within the company.” he said.

Also read: Tata Communications unfazed by US chip export curbs; in talks with enterprises for NVIDIA-powered data centre: MD