India AI Digest — Friday, August 7, 2026

• TCS is putting up to $7B behind owned AI data-centre capacity through a TPG joint venture; HCLTech is making the same bet at smaller scale; Infosys's CEO says the board reviewed and rejected the business.

---

SERVICES · COMPUTE · STRATEGY · August 5, 2026

TCS commits up to $7B to owned AI compute; Infosys's board declines the business

TCS has structured HyperVault, an AI data-centre joint venture with TPG, with TCS holding 51%. TPG has put in $1 billion; combined partner equity is $2 billion, with a further $4.5–5 billion in debt targeted to fund the buildout. Land is secured in Pune and Andhra Pradesh. Per Analytics India Magazine's reporting, TCS plans $6–7 billion in total investment over five to seven years, targeting 1 gigawatt of capacity at a cost ratio of roughly $1 billion per 150 MW. CEO K Krithivasan frames the move as part of a "full-stack" ambition spanning chip design, infrastructure, models, platforms, and AI agents. HCLTech is running the same play at a smaller scale — its ₹3,500 crore (~$368 million), up-to-50MW data-centre commitment, announced July 13, is now being framed by corporate VP Sriram Hariharan around the revenue math: colocation earns "roughly one unit of revenue per square foot," against a 10–15x multiple for full-stack AI hosting. CEO C Vijayakumar's framing is owning AI capacity rather than renting it. Infosys has taken the opposite position. CEO Salil Parekh said the company "reviewed it with its management team and board, and had decided against making any move into that business at this stage," pointing instead to an asset-light path — the Anthropic/Topaz tie-up, an ExxonMobil immersion-cooling partnership, and the Topaz Fabric multi-model workload platform.

This account traces to a single Analytics India Magazine report aggregating statements attributed to all three companies; none of the specific TPG/HyperVault deal terms are independently confirmed by a TCS release in what's available here, so treat the dollar figures as source-conditional pending primary confirmation.

What this means. Three of India's largest IT services firms have now taken three different positions on the same structural question — should a services company own the compute layer under its AI offering, or buy it as a service — and the split runs along a coherent line: appetite for capital risk versus adjacency to the existing business model.

TCS's number is the one worth sitting with. $6–7 billion is not a services-firm-scale AI capex line; it's an infrastructure investor's number, financed accordingly — majority-owned JV, external equity partner, and a debt raise nearly twice the combined equity. That structure spreads the capital risk but also means TCS is underwriting utilization risk across a partner's balance sheet, not just its own. HCLTech's number is a tenth the size and already committed as of July, so today's framing is HCLTech restating the economics case rather than a new decision. Infosys's move is the one with the clearest logic and the most exposure if the logic is wrong: asset-light avoids the utilization risk entirely, but it also means Infosys has no compute-margin upside if owned AI infrastructure turns out to be where SI-layer differentiation actually lives over the next few years. Parekh's framing — reviewed and rejected, not simply deferred — reads as a considered no rather than a wait-and-see.

India angle. For the SI layer, this is now a live strategy fork rather than a shared industry direction. Two of the three largest listed IT services companies are committing multi-billion-dollar capital to domestic AI data-centre capacity; the third is explicitly not. Whichever position proves right will reset how the sector is valued on AI exposure — investors currently have three different company-defined AI-revenue metrics from this same earnings season and now three different infrastructure postures layered on top, with no common yardstick across either axis. For India's compute-infrastructure buildout more broadly, TCS's 1GW target — if it delivers on the stated timeline — would be a meaningful addition to the domestic AI-compute base, in the same conversation as AirTrunk-Blackstone's 5GW plan and HCLTech's own Odisha commitment, though all three remain pre-delivery.

Behind the news. HCLTech's ₹3,500 crore, 50MW commitment was first covered on July 16, the same week it disclosed a $171M advanced-AI revenue line; it followed that with a $1.5B Odisha data-centre commitment with Sarvam AI on July 25 and a GPU scale-up disclosure covered on July 31. Infosys, by contrast, used its own Q1 FY27 print — covered July 24 — to disclose an 8.2% AI-revenue share while trimming FY27 growth guidance, with no infrastructure-ownership component in that report. Today's item is the first time the three companies' positions have been placed against each other directly, and the first appearance of TCS's HyperVault/TPG structure in this archive.

What to watch. TCS's targeted $4.5–5 billion debt raise for HyperVault — whether it closes, and on what terms, is the first checkpoint on whether the $6–7B figure is financeable as stated. Also watch Infosys's Q2 FY27 call in October for whether the asset-light position holds if TCS or HCLTech report utilization gains from owned capacity in the interim.

Source: Analytics India Magazine, "TCS, HCLTech Think Data Centres are the Next IT Frontier. Infosys Disagrees." → link

Confidence: Medium. Single secondary source aggregating statements from all three companies; TPG/HyperVault deal terms not independently confirmed against a TCS primary release.

---

Position movements