Sify, one of India’s biggest data centre providers, announced on Tuesday a new pay-per-use pricing structure for its NVIDIA-certified AI-ready data centres across the country. The company now offers an hourly pricing model that includes hosting, power, and infrastructure costs.
The company stated this is an attempt to eliminate the entry-cost barrier and fixed-cost infrastructure risk. Initially, Sify will support NVIDIA GPUs, including the H100, H200, B200, GB200 NVL2, and GB300 NVL72 platforms.
Sify has AI-ready NVIDIA DGX-certified data centres in Mumbai, Chennai, and Noida, and it recently unveiled one of India’s largest AI-ready data centres in Siruseri, Chennai. NVIDIA’s DGX certification ensures that data centre infrastructure, software platforms, and managed services are fully optimised for NVIDIA DGX AI systems, turnkey solutions that integrate high‑performance GPUs, storage, and software.
Pay Only For GPU Utilisation
In a conversation with AIM, Sharad Agarwal, CEO of Sify Infinit Spaces, said, “While there is a lot of excitement around the adoption of AI, our observation was also that the pace at which these adoptions should happen in India has not been to the extent it is in, let’s say, the United States.”
To help achieve this goal, he added that the company wanted to address the cost effectiveness and offer a variable pricing model that might be attractive for cloud companies to set up shops in India.
To begin with, AI applications interact with GPUs located in data centres through the cloud to carry out tasks.
“India currently has local AI cloud companies offering GPUs as a service. We do not yet have the big GPU as a service providers from the US setting up shop here, and they are interested in coming here,” added Agarwal.
Consequently, Sify’s new pricing model has been developed so that clients are not required to pay a fixed cost for space when deploying their racks. Instead, charges apply exclusively based on the GPU service. Clients will only incur costs when GPUs are in use and power is activated, with options for hourly, monthly, or yearly pricing.
“So that way, the pricing model is different — not based on the space, not on the power, but only on the utilisation of the GPUs,” he added. For context, colocation data centres present different pricing models for clients, primarily based on the space utilised within the data centre, power consumption, or allocated power.
Agarwal noted that while foreign companies do not wish to overlook India in the AI ecosystem, they often express concerns about its unique challenges. Specifically, they worry about the difficulties in developing, building, and renting infrastructure.
The company also claims that its competitors do not offer a similar pricing model and that they continue to use the traditional approach. For example, Yotta, a provider of GPU-as-a-service in India, features a rack-based pricing model on its website, offering five racks for ₹99,999.
Why India Fits the Bill
Recent advancements in India’s AI and data landscape indicate a growing need to host AI workloads and GPUs.
The country’s ₹10,000 crore-plus IndiaAI Mission is poised to develop domestically created AI models, having recently chosen Sarvam AI to build one such model. Once the project is complete, there will be a heightened demand to scale and deploy the AI model using GPUs in data centres.
Although Sify promotes a distinct pricing model, numerous other companies catering to AI workloads with GPUs are actively investing in and enhancing their services, particularly through collaborations with the government.
The IndiaAI mission seeks to deliver over 18,000 GPUs via public-private partnerships involving firms such as Jio Platforms, NxtGen Data Centre, Locuz Enterprise, E2E Networks, CtrlS DataCenters, CMS Computers, Orient Technologies, Tata Communications, Vensysco, and Yotta Data Services.
While Sify does not appear in the initial list of bidders for IndiaAI Mission’s AI compute, when asked, the company declined to comment, stating that discussions are in progress and that the infrastructure is ready.
Furthermore, global companies are increasingly choosing India to manage their AI workloads. For instance, OpenAI announced that it will permit organisations within the country using OpenAI’s APIs, as well as ChatGPT Enterprise and Edu, to store customer content in India at rest. Given that India boasts the second-largest user base of AI products, the country presents an opportunity for global firms to deploy GPUs within the country.
Besides, the country’s Data Privacy and Data Protection Rules (DPDP), currently in their draft stage, propose that entities handling certain kinds of personal data must store and process that data within India.
Furthermore, the already booming AI startup ecosystem will drive the GPU services market, primarily if it aims to build on open-source AI models and host its products at scale.
Consequently, companies that allow the hosting of AI workloads and GPUs in their data centres will benefit all parties, including the government, sovereign AI mission, foreign entities, and Indian AI startups.
