Increasing competition, new players expected to lower GPU prices: NxtGen's CEO

The Ministry of Electronics and IT (MeitY) launched the AI Compute Portal under the ₹10,000 crore India AI Mission, last week. This platform will provide access to 18,000 graphic processing units (GPUs), a specialized chip that is considered best fit for AI workloads. Through this portal, students, startups, researchers, and government departments will be able to access some of the high-performance GPUs.

Bengaluru-headquartered NxtGen Datacenter and Cloud Technologies (NxtGen) is one of the service providers that has been empanelled by the government to provide GPUs to the AI mission. NxtGen’s chief executive officer (CEO) and managing director AS Rajgopal told TechCircle in an interview that NxtGen is offering some of the most advanced GPUs including Nvidia’s H100 and H200, and AMD’s MI300X towards the initiative. 

“India’s AI mission is one of the most comprehensive initiatives of its kind globally. This mission aims at encouraging AI adoption by subsidising up to 40% of the cost incurred by academic institutions, research organizations, and small and medium enterprises (SMEs),” he said. Rajgpal added that the company is charging ₹140 per GPU hour, lower than the global benchmark of $2.5 (~₹218).

“With the launch of a dedicated portal, NxtGen’s system is now fully integrated with the government’s AI mission. Once a project is approved, workloads are automatically provisioned on NxtGen’s cloud infrastructure.”

AI infrastructure is capital-intensive, and the cost of high-performance GPUs alone is substantial. Apart from data center investments and backend setup, NxtGen requires at least $400 million solely for GPU servers, said Rajgopal. The company aims to procure about 12,000 GPUs more. “We estimate that over the next two to three years, it will need between $2 billion to $3 billion to scale AI cloud operations effectively.”

GPU shortage and market monopoly

ChatGPT-maker OpenAI released its latest large language model (LLM) GPT-4.5 last month. At the time of its launch, OpenAI head and co-founder Sam Altman highlighted the issue of GPU shortage. “We’ve been growing a lot and are out of GPUs,” he wrote in an X post. “We will add tens of thousands of GPUs next week and roll it out to the Plus tier then … This isn’t how we want to operate, but it’s hard to perfectly predict growth surges that lead to GPU shortages.” This is not the first time Altman has spoken about it; in the past he said that the lack of compute resources has hindered the company’s ability to roll out models faster.

Currently, Nvidia holds the market monopoly in GPUs. Riding on the generative AI wave, which is highly dependent on these specialised chips, Nvidia became the most valuable company in June 2024. The company's market capitalization surpassed $3.3 trillion, overtaking giants like Microsoft and Apple.

Nvidia’s chips also dominate among the 10,000 GPUs being offered as part of MeitY’s AI Compute Platform. Commenting on this, Rajgopal said, “Currently, Nvidia enjoys a near-monopoly in the AI GPU space, making price negotiations almost impossible. However, competition is growing, and as more companies enter the market, prices are expected to decline.”

Besides leading AI chips providers Nvidia and AMD, NxtGen is also working with Santa Clara-based d-Matrix. “The reliance on high-cost GPUs can be reduced by building smaller, optimized models. This is the key advantage of inference, which allows AI to operate without the heavy compute demands of training. d-Matrix specialises in inference-specific GPUs. These GPUs are significantly more cost-effective compared to Nvidia’s premium models, which dominate the market.”

NxtGen itself is building a model to serve enterprises’ financial queries. This model is based on Meta’s Llama 3 open source model.

“Currently, enterprises struggle to integrate tools like ChatGPT into their workflows beyond basic tasks like drafting emails, which often lose appeal after a few days. Instead, we are focused on building AI applications that directly contribute to business efficiency. Our financial AI assistant will be the first of many such innovations.”

Scaling to more than 600 edge data centers

Currently, NxtGen operates five major data centers across India—in Bangalore,Mumbai, and Ahmedabad. Beyond these, NxtGen has edge data centers in 36 locations, including regions like Belgium, Raipur, and Puri. The company’s long-term vision is to establish AI and cloud services in every district headquarters, with a goal of reaching 700 edge data centers across India. As part of this expansion, NxtGen is also hiring over 400 engineers to scale services both in India and internationally. 

Further, Rajgopal is actively advocating for the government to designate data center parks outside cities rather than within them. “Building data centers in urban areas is not sustainable for two key reasons. First, real estate costs skyrocket, making large-scale infrastructure investments far more expensive. Second, urban power grids are already under strain, and AI-driven data centers demand exponentially more electricity.” Notably, data centers hosting AI workloads require up to 10 times more power.

He said that globally, this practice is prevalent. “India should adopt a similar approach by creating dedicated data center parks with at least a gigawatt of connectivity. Developing three to four such parks across the country would be a far more efficient and sustainable solution than treating data center expansion as just another real estate business.”