HCLTech bets $1.48 billion on AI infrastructure in Odisha with Sarvam as its model partner HCLTech will invest ₹142.57 billion ($1.48 billion) to build its first AI data center in Bhubaneswar, Odisha, with Sarvam AI as its model partner and the Odisha government providing financial assistance. The project, part of the Odisha Sovereign AI Park, aims to deliver sector-specific AI applications to private and government clients, combining HCLTech's full-stack AI capabilities with Sarvam's foundation models. HCLTech also plans a global technology center in Bhubaneswar expected to house 5,000 employees by 2028. HCLTech is putting real money behind sovereign AI in India, with a ₹14,257 crore data center plan in Bhubaneswar and Sarvam sitting at the model layer. HCLTech spent decades making other companies' infrastructure work. Now it wants to own part of the stack itself. On Friday, Reuters reported that India's third-largest IT services firm will invest ₹142.57 billion, about $1.48 billion, to set up its first AI data center in Odisha with Sarvam AI and the state government. The site is planned for Bhubaneswar, at the upcoming Odisha Sovereign AI Park. That location matters. You don't put this kind of money into a data center because AI services sound fashionable. You do it because the old services model is getting squeezed, and because clients now want compute, models, applications and compliance wrapped together instead of passed through four separate vendors. HCLTech has decided that running someone else's infrastructure is no longer enough. HCLTech is moving up the stack The project will use HCLTech's full-stack AI capabilities and Sarvam's foundation models to offer sector-specific AI applications to private companies and government customers, according to the company filing cited by Business Standard. The same report said the capital outlay includes financial assistance from the Odisha government. That is not a small detail. Sovereign AI only becomes real when governments are willing to help pay for the land, power and the institutional backing that heavy compute needs. There is another piece in Bhubaneswar. HCLTech is also opening a global technology center in the city that is expected to house 5,000 employees and start operations by 2028, Reuters reported through The Economic Times. Earlier reporting from Business Standard said HCLTech had planned about ₹730 crore for an AI-optimised data center and global development center in Bhubaneswar, while the company's broader data center push was routed through new subsidiaries and could scale to 50 megawatts of capacity. That gives the Odisha plan a more concrete shape: not just a server hall, but an operating base. Odisha has been courting exactly this kind of investment. Chief Minister Mohan Charan Majhi visited HCLTech's Noida campus in April 2025 and publicly invited the company to set up in Bhubaneswar, according to ANI reporting carried by ThePrint. On Friday, the MoU was signed in the presence of Majhi, HCLTech chairperson Roshni Nadar Malhotra and CEO C Vijayakumar, Business Standard reported. Those names make the announcement harder to dismiss as a routine expansion note. The uncomfortable point for Indian IT is simple. Asset-light outsourcing was the playbook for forty years. AI infrastructure is asset-heavy. Data centers need power, cooling, chips, land, capital and patience. HCLTech is choosing that burden because the margin pool in AI may sit closer to compute and models than to plain managed services. That is a real break with habit. Sarvam gives the project its reason Sarvam is not decoration here. The Bengaluru company raised $234 million in June in the first close of a planned $300 million Series B at a $1.5 billion valuation, according to HCLTech's own release and TechCrunch. HCLTech led the round with a $150 million investment for a 10.46% stake. Bessemer Venture Partners joined the round, with Khosla Ventures and Peak XV Partners returning. That made Sarvam India's newest AI unicorn, but the label is less important than the role it now plays. TechCrunch reported at the time that the plan was to combine Sarvam's models with HCLTech's enterprise relationships, engineering workforce and software assets to build AI products for businesses and governments. The Odisha data center turns that strategy from a funding-story promise into infrastructure. Sarvam supplies the model layer. HCLTech brings the sales reach and delivery muscle, along with the managed services wrapper. The state supplies the sovereign AI setting. Sarvam's language work is the practical reason this partnership makes sense. Its own model documentation says coverage varies by product, but its speech-to-text, translation and document intelligence tools cover 22 Indian languages plus English. Its larger chat models cover fewer languages. That distinction matters, because government services in India don't run only in English, and a model that can process Odia documents, Hindi speech or Bengali text is more useful to a state department than a generic English-first chatbot with a local wrapper. Frankly, that is the bet. HCLTech is not saying it can outspend Microsoft, Amazon or Google on global cloud infrastructure. It is saying Indian enterprises and public agencies will pay for AI that runs closer to home, respects sovereignty requirements and understands local languages well enough to be used in production. If that assumption is right, Bhubaneswar becomes more than a delivery center. It becomes part of the product. There are still open questions. No enterprise customer list has been announced for the full-stack AI offering, and data centers are hard businesses even before you add GPU supply, power costs and utilisation risk. Tata Consultancy Services has already announced a $2 billion AI data center plan with TPG, while Business Standard reported that Infosys CEO Salil Parekh has ruled out entering the segment. HCLTech is not alone, but it is taking a clearer ownership position than most of its services peers. That is why this announcement matters. A company built on managing infrastructure has decided the AI cycle requires owning some of it. You should watch whether customers follow, because the Bhubaneswar site only proves the thesis if HCLTech can turn sovereign compute and Sarvam's models into contracts that are larger, stickier and harder to displace than the outsourcing work it grew up on. Also read: Anthropic upgrades Claude voice mode to run Opus and Sonnet while automating Gmail, Calendar, and Slack https://startupfortune.com/anthropic-upgrades-claude-voice-mode-to-run-opus-and-sonnet-while-automating-gmail-calendar-and-slack/ • Midjourney buys Co-Star to seed a consumer app empire it never needed to build before https://startupfortune.com/midjourney-buys-co-star-to-seed-a-consumer-app-empire-it-never-needed-to-build-before/ • CATL's energy storage business grew 88% in six months as AI data centers become its next major market https://startupfortune.com/catls-energy-storage-business-grew-88-in-six-months-as-ai-data-centers-become-its-next-major-market/