Verdict: TCS and HCLTech are making multi-billion-dollar bets that owning AI data center infrastructure — not just writing software for it — is the next growth frontier for Indian IT services. TCS has secured $1 billion from TPG and plans up to $6.5 billion total for 1 GW of AI-ready capacity through its HyperVault joint venture. HCLTech committed ₹3,500 crore ($368 million) for 50 MW of capacity and followed it with a ₹14,257 crore ($1.48 billion) sovereign AI data centre in Odisha. Infosys has looked at the same opportunity and walked away, betting instead on AI software partnerships with Anthropic and ExxonMobil rather than owning servers. Both approaches can work — but they serve fundamentally different theories of where the value in AI infrastructure will accrue.
Last verified: 2026-08-07 · TCS HyperVault: up to $6.5B for 1 GW · HCLTech: ₹3,500 Cr + ₹14,257 Cr across two projects · Infosys: no data centre investment, pursuing asset-light AI partnerships · Pricing and scope are volatile — re-check monthly.
Why Are Indian IT Companies Investing in AI Data Centers?
Indian IT services firms traditionally wrote software that ran on someone else's servers — typically hyperscaler cloud from AWS, Azure, or Google Cloud. The shift toward owning the underlying compute infrastructure is driven by three converging forces: AI workloads demand purpose-built, liquid-cooled, GPU-dense facilities that standard colocation cannot serve; India's government is actively subsidizing sovereign AI infrastructure through the IndiaAI Mission; and full-stack AI services — infrastructure plus models plus applications — can generate 10–15 times the revenue per megawatt compared to traditional colocation, according to HCLTech corporate VP Shriram Haran.
The IndiaAI Mission has already deployed over 38,000 GPUs as of mid-2026, with a government target of 100,000 publicly accessible GPUs by December 2026, according to IndiaAI Mission CEO Abhishek Singh. India's total data centre capacity sits at roughly 1.5–2.2 GW in 2025 and is projected to reach anywhere from 6.5 to 12 GW by 2030, depending on the analyst estimate — Wood Mackenzie projects 12 GW, while Jefferies and Deloitte forecast 8–10 GW.
This capacity expansion represents a multi-billion-dollar investment opportunity. India's data centre market has attracted nearly $94 billion in investments since 2019, according to TCS's own November 2025 press release.
What Is TCS's HyperVault Data Center Strategy?
TCS (Tata Consultancy Services) announced its HyperVault AI data center business on November 20, 2025, securing a $1 billion investment from TPG — the global alternative asset management firm — through TPG Terabyte, facilitated via TPG Rise Climate and its Global South Initiative. The joint venture commits up to ₹18,000 crore (approximately $2.1 billion) in combined equity, with TPG investing up to ₹8,820 crore and holding a final shareholding between 27.5% and 49%. TCS retains majority control with 51% equity and the right to appoint the majority of HyperVault's board directors.
CEO K. Krithivasan outlined the scale: HyperVault targets over 1 GW of AI-ready data center capacity over 5–7 years, with roughly $1 billion of investment per 150 MW — translating to a total build cost exceeding $6.5 billion. The financing mix combines equity from TCS and TPG with $4.5–5 billion in projected debt. TCS has secured land in Pune and Andhra Pradesh for the buildout.
Krithivasan frames HyperVault as part of a "full-stack ambition" to position TCS as an AI-first operator. The target customers include AI providers, deeptech companies, hyperscalers, and — notably — government entities in India and Indian enterprises. The facilities will offer purpose-built, liquid-cooled data centers with high rack densities designed for GPU-intensive AI training and inference workloads.
What Is HCLTech's AI Data Center Investment Plan?
HCLTech approved an investment of up to ₹3,500 crore ($368 million) on July 13, 2026, to establish AI data centers in India through wholly owned subsidiaries, targeting up to 50 MW of installed capacity. The board approved the investment alongside HCLTech's Q1 FY2027 results — quarterly revenue of ₹34,579 crore (up 13.9% year-on-year) and net income of ₹4,624 crore (up 20.3%).
CEO C. Vijayakumar positioned the move as essential to becoming a "full-stack AI technology solutions provider," telling analysts that owning AI capacity rather than renting it represents the bigger commercial opportunity. The investment covers AI infrastructure, GPU-intensive training and inference workloads, cloud operations, DevOps, and enterprise AI solutions.
The Odisha Sovereign AI Park Project
On July 24, 2026, HCLTech announced a separate, larger commitment: a ₹14,257 crore ($1.48 billion) investment to build its first AI data centre at the Odisha Sovereign AI Park in Bhubaneswar, in partnership with Indian AI startup Sarvam and the Government of Odisha. The project includes state government financial assistance and was announced in the presence of Odisha Chief Minister Mohan Charan Majhi.
The Odisha facility's phased go-live target is 2028. It will combine HCLTech's AI infrastructure capabilities with Sarvam's foundation models to deliver sector-specific AI applications, multilingual AI services, and sovereign-grade compute for both private and public sector workloads. Chairperson Roshni Nadar Malhotra framed the project as aligning with Digital India and Atmanirbhar Bharat. The partnership also builds on HCLTech's earlier $150 million investment in Sarvam AI.
Why Is Infosys Not Investing in Data Centers?
Infosys has explicitly decided against entering the AI data center infrastructure business. During a recent earnings call, CEO Salil Parekh stated that Infosys had reviewed the data center opportunity with its management team and board, and concluded that it should not make any move into that business at this stage.
Instead, Infosys is pursuing an asset-light strategy built on AI partnerships and platform integration:
Anthropic partnership (February 17, 2026): Infosys and Anthropic announced a collaboration to integrate Claude models and Claude Code with Infosys Topaz, building enterprise-grade AI agents for telecommunications, financial services, manufacturing, and software development. A dedicated Anthropic Center of Excellence within Infosys launched with telecom as the initial focus.
ExxonMobil immersion cooling partnership (February 12, 2026): Infosys expanded its collaboration with ExxonMobil to deploy liquid immersion cooling for AI data centers. ExxonMobil supplies its Data Center Immersion Fluids (dielectric, non-conductive), while Infosys Topaz handles real-time AI-driven cooling optimization and Infosys Cobalt manages cloud and hybrid deployment. The partnership targets PUE as low as 1.03–1.08 versus 1.50+ for traditional air cooling.
Topaz AI platform: Infosys's AI-first services framework integrates generative AI, cloud modernization, and automation — sold to enterprises without Infosys needing to own the physical infrastructure underneath.
The strategic logic: most enterprise customers will continue relying on hyperscalers (Google, Microsoft, Amazon) for raw compute infrastructure, and an IT services firm's edge lies in software and integration — not in operating power-hungry facilities.
TCS vs HCLTech vs Infosys: AI Data Center Strategy Comparison
| Dimension | TCS (HyperVault) | HCLTech | Infosys |
|---|---|---|---|
| Investment committed | Up to $6.5B (equity + debt) | ₹3,500 Cr + ₹14,257 Cr (~$1.85B total) | $0 (no data center investment) |
| Target capacity | 1+ GW over 5–7 years | 50 MW (initial) + Odisha sovereign facility | N/A |
| Funding structure | 51:49 JV with TPG ($1B equity) + $4.5–5B debt | Wholly owned subsidiaries + Odisha state assistance | N/A |
| Key partnerships | TPG ( equity partner) | Sarvam AI, Odisha government | Anthropic, ExxonMobil, NVIDIA, Google Cloud |
| Strategic thesis | Own infrastructure, sell to AI providers and government | Full-stack AI from infrastructure to applications | Asset-light: own software and integration, rent compute |
| Geographic focus | India (Pune, Andhra Pradesh) | India (Odisha, pan-India subsidiaries) | Global (partner-led) |
| Target customers | Hyperscalers, AI companies, government, enterprises | Enterprises, government, sovereign workloads | Enterprises in telecom, financial services, manufacturing |
What Are the Risks of IT Companies Owning Data Centers?
The bear case for TCS and HCLTech's infrastructure bet centers on capital intensity and returns. Three concerns stand out:
1. Lower returns than core business. Brokerage BofA Capital (Bob Caps) called TCS's data center plan a "negative surprise," citing lower expected returns than its core IT services business. Data centers are capital-heavy infrastructure projects with long gestation periods — typically 3–5 years before meaningful revenue contribution — unlike software services where revenue scales with headcount.
2. AI infrastructure may commoditize. If hyperscaler cloud prices for GPU compute continue falling — as they have through 2025–2026 — the margin advantage of owning infrastructure could compress. Former SBI and HSBC CIO Aparna Kumar estimates that a meaningful data center entry could cost $5–10 billion and argues most enterprises will continue relying on hyperscalers rather than Indian IT firms for raw compute.
3. No structural edge in operations. Running a data centre at scale requires expertise in power procurement, cooling engineering, physical security, and network operations — capabilities that traditional IT services firms have not historically owned. Reliance, Adani, NTT DATA, and specialized operators (AdaniConnex has announced a 2.6 GW pipeline) are already competing in the Indian market.
Conversely, the risk for Infosys's asset-light approach is leaving significant revenue on the table if full-stack AI infrastructure services become the dominant procurement model and clients want one partner owning the entire stack.
What Does India's AI Sovereignty Push Mean for This Split?
The Indian government's AI sovereignty agenda is a tailwind for TCS and HCLTech's infrastructure bets. The IndiaAI Mission (total budget: ₹10,371.92 crore / ~$1.25 billion) has deployed over 38,000 GPUs as of mid-2026 — nearly four times the original 10,000-GPU target — accessible to startups, researchers, and government agencies at a subsidized rate of ₹65 per GPU-hour (roughly 60% below commercial cloud pricing). The target is 100,000 publicly accessible GPUs by December 2026.
Data centers have received formal "infrastructure status" in India, unlocking access to long-tenure project financing. The government has also offered a 20-year tax holiday for hyperscalers using Indian data centres to service global clients, and the National Data Centre Policy framework is being updated to streamline facility development and grid connection.
For HCLTech's Odisha project, the state government partnership explicitly frames the data centre as sovereign AI infrastructure — compute owned and operated on Indian soil for Indian workloads, reducing dependence on overseas infrastructure. This sovereignty framing gives TCS and HCLTech a regulatory and geopolitical rationale that pure commercial data centre operators do not have.
What This Means for You
If you're an enterprise evaluating AI infrastructure in India: The TCS-HCLTech data center buildout means more domestic options for GPU-intensive AI workloads, potentially with better data residency and compliance terms than renting from hyperscalers. But these facilities won't be fully operational until 2028 at earliest — plan for hyperscaler dependency in the interim.
If you're an investor in Indian IT: The split reveals a genuine strategic fork. TCS and HCLTech are converting their services margins into infrastructure capex — a higher-risk, higher-reward bet. Infosys is protecting its asset-light margins and betting partnerships, not capital, will capture the AI services market. Watch whether full-stack AI infrastructure procurement actually wins mindshare over the next 2–3 years.
If you're building AI products for the Indian market: The IndiaAI Mission's 38,000+ GPUs at ₹65/GPU-hour represent the cheapest compute path for eligible startups and researchers. The TCS-HCLTech private facilities will serve enterprises and government — not startups — at commercial rates when they come online.
How Does India's Data Center Pipeline Compare Globally?
India's data center market is moving from roughly 1.5–2.2 GW in 2025 toward 6.5–12 GW by 2030, representing a compound annual growth rate of approximately 40% per Wood Mackenzie. AI-dedicated capacity specifically is expected to grow nearly 24-fold — from 275 MW to 6,546 MW by 2030.
This buildout is attracting significant capital: nearly $94 billion in cumulative investment since 2019, with Deloitte projecting India may attract $200 billion in data centre investment by 2030 as part of a broader $800 billion Asia Pacific investment pipeline. For context, this positions India alongside — though behind — the United States and China in AI infrastructure scale, but with structurally lower construction costs, land rates, and power tariffs as competitive advantages.
The broader AI infrastructure spending context matters too — Big Tech's collective $750 billion AI capex binge faces its own scrutiny around whether hardware depreciation will outrun revenue. Indian IT firms entering this space must navigate the same depreciation dynamics.
The push also sits within a broader reshaping of India's $315 billion IT industry by AI, where automation is simultaneously compressing traditional services pricing and opening new infrastructure-adjacent revenue streams. The July 2026 Indian IT stock rally of 22% partially reflected investor optimism about this transition, though analysts remain divided on whether the rally reflects fundamentals or positioning.
FAQ
Q: How much is TCS investing in AI data centers?
A: TCS plans up to $6.5 billion total for over 1 GW of AI-ready data center capacity through its HyperVault joint venture with TPG. TPG committed $1 billion in equity (up to ₹8,820 crore), and TCS holds 51% with plans to raise $4.5–5 billion in additional debt. The build targets 5–7 years with sites in Pune and Andhra Pradesh.
Q: How much is HCLTech investing in AI data centers?
A: HCLTech has two committed investments: ₹3,500 crore ($368 million) for up to 50 MW of AI data center capacity approved July 13, 2026, and a ₹14,257 crore ($1.48 billion) sovereign AI data centre in Odisha announced July 24, 2026, in partnership with Sarvam AI and the Odisha government. Combined, the total exceeds $1.85 billion.
Q: Why is Infosys not investing in data centers?
A: Infosys CEO Salil Parekh stated the company reviewed the data center opportunity with its board and decided against entering the business at this stage. Infosys prefers an asset-light approach — partnering with Anthropic (Claude models integrated into Infosys Topaz), ExxonMobil (immersion cooling technology), and hyperscalers — betting that its software and integration capabilities, not infrastructure ownership, will capture AI services value.
Q: How many GPUs does India's IndiaAI Mission have?
A: The IndiaAI Mission has deployed over 38,000 GPUs as of mid-2026, up from a 10,000-GPU original target, accessible at a subsidized rate of approximately ₹65 per GPU-hour. The government's stated goal is 100,000 publicly accessible GPUs by December 2026, with a third procurement tender currently adding approximately 3,850 units including 1,050 Google Trillium TPUs.
Q: What is the revenue advantage of owning AI data centers vs renting?
A: HCLTech corporate VP Shriram Haran told Analytics India Magazine that traditional colocation earns roughly one unit of revenue per square foot or megawatt, while a full-stack AI offering can push that 10–15 times higher with better returns on capital. This revenue multiplication is the core economic rationale for IT services firms owning infrastructure rather than just reselling hyperscaler capacity.
Q: Will enterprises prefer Indian IT firm data centers over hyperscalers?
A: Not necessarily. Former SBI and HSBC CIO Aparna Kumar estimates a meaningful data center entry could cost $5–10 billion and argues most enterprises will continue relying on hyperscalers (Google, Microsoft, Amazon). The TCS-HCLTech bet is that sovereignty requirements, full-stack integration, and government demand — not enterprise preference — will be the differentiating demand driver.
Every claim here is traced to a primary source, dated, and listed under Sources. Research and drafting are AI-assisted; editing, verification and publication are human decisions, and a person is accountable for what appears on this page. How we work →
Discussion
0 comments