Yotta Closes Largest Single India AI Funding Round
Yotta Data Services, the Hiranandani Group infrastructure spin-out, has closed a USD 500 million funding round led by Macquarie Asset Management with participation from Temasek. The capital, announced on April 25, anchors a 24,000 GPU AI supercluster in Greater Noida and is the largest single commitment to Indian AI infrastructure on record.
The round closes a gap that has nagged Indian sovereign AI ambitions since the IndiaAI Mission was approved in 2024. Domestic compute supply has consistently lagged demand from Indian language model labs, public sector pilots, and global hyperscalers seeking peering capacity inside the country.
What The Round Buys
The Greater Noida campus is engineered for 100 MW of usable IT load in phase one, with provisions to expand to 250 MW by 2028. The first 8,000 NVIDIA H100-class systems are already racked, and the balance is scheduled to ship through the third quarter on Foxconn-built rack designs sourced from Taiwan.
Yotta has prioritised liquid cooling for the first deployment to support GB200 NVL72 racks for hyperscaler tenants in 2027. Power conditioning has been built around two redundant 220 kV substations with the Uttar Pradesh Power Corporation, a faster grid connection than any prior Indian build.
The operating profile is also unusual for India. Yotta is targeting a power usage effectiveness below 1.30, well ahead of the national average above 1.55, and is contracting renewable supply through a JSW Energy green PPA covering roughly 80% of phase one demand.
Why Macquarie And Temasek Wrote The Cheques
Macquarie's Asia Pacific infrastructure unit has been quietly building a thesis around regional AI campuses for two years, and the Yotta cheque follows an earlier Australian Pacific Connect undersea cable investment that fits the same compute logistics story. For Macquarie, India is the only emerging Asia market with both demand density and a credible regulatory floor.
Temasek's involvement is the more strategic signal. Singapore's sovereign investor has been more selective with Indian tech bets than its India-focused peers and tends to back assets that complement Singapore's own digital infrastructure base. Indian compute capacity reduces the latency and policy friction Singapore faces when serving Indian customers from Changi or Tuas data centres, and Temasek now sits on both sides of that flow.
How It Reshapes Asian AI Compute
The Yotta deal lands in a regional market where Singapore, Malaysia, Indonesia, and Australia have each been positioning to absorb hyperscaler workloads. India's structural advantage is power cost and demand size; its structural weakness has been speed of build. By combining IndiaAI Mission incentives with sovereign equity from Temasek, Yotta gets a 12 to 18 month time advantage over rivals raising on commercial terms alone.
For Singapore-based hyperscalers, the implication is that Indian sovereign workloads can now stay in India. That may relieve power pressure on Tuas and Loyang in 2027, but it also costs Singapore some of the revenue that had been booked as cross-border. Australian operators including NextDC and AirTrunk face a different challenge: their core South Asia hedge depended on India's capacity gap persisting longer.
What MeitY And IndiaAI Mission Get
MeitY's approval band gives Yotta access to roughly USD 350 million in matched compute incentives over six years, contingent on capacity availability for IndiaAI Mission labs and Bhashini language model partners. About 60% of contracted capacity is now reserved for Indian language model startups, public sector workloads, and government research, with the rest allocated to commercial peering.
The political messaging is straightforward. Minister Ashwini Vaishnaw has repeatedly framed India's AI bet as built domestically rather than rented from US or Chinese hyperscalers, and Yotta is now the most concrete evidence of that pitch. Sarvam AI and Krutrim, two of the highest profile Indian language model labs, have both signalled they will move workloads to Greater Noida from foreign capacity over the next two quarters.
Risks Investors Will Watch
Three risks dominate the next twelve months. First, GPU delivery slippage; even with Foxconn rack priority, India ports have lengthened lead times by an average of 11 weeks during 2025 customs reform. Second, grid stability; Uttar Pradesh has had two notable industrial brownouts in the past year, and AI training intolerance to power events is high. Third, currency; the rupee depreciation pace has tightened operating margins for prior data centre rounds, and the Yotta financing is partly in US dollars.
The deal's structure absorbs some of this. Macquarie and Temasek have both written long duration tranches, and Yotta has retained its option to bring in a strategic operator stake later in 2026 if specific delivery milestones slip.
What Comes Next
Watch the Q3 2026 capacity test. If the Greater Noida site can deliver the contracted 16,000 GPU live throughput with sub-five hour fault recovery to its first three anchor tenants, the round will read as a category opener for Indian AI infrastructure. If delivery slips beyond Q4, the next India compute headline is likely to come from a rival site, with Reliance Jio Platforms and Adani Connex both holding shovel ready land in Pune and Hyderabad.
Indian Sovereign AI Implications
The broader Indian sovereign AI agenda gets a clear data point from this round. MeitY has been arguing for two years that domestic compute is a precondition for sovereign model development; until now, the most visible proof points were Bhashini language tooling and Krutrim's launch announcements. Yotta's funded campus turns that argument into physical capacity that ministers can show foreign partners and that startups can actually book.
For regional alignment, the deal also signals a closer Singapore India compute corridor. Temasek's stake gives Singapore both equity exposure to Indian AI growth and policy visibility into how Indian sovereign workloads are routed. That can either deepen Singapore's position as the regional AI hub or, in time, shift some workloads inland to where the demand and incentive base sits. Either outcome is healthier for the region than the recent stretch in which Singapore was the only credible operator.
Frequently Asked Questions
Does the Yotta round include any government equity?
No. The IndiaAI Mission contribution is structured as compute incentive credit and matched grants, not equity. The capital structure is private at the campus level, with Macquarie and Temasek the dominant new shareholders.
Will Yotta serve foreign hyperscalers?
Yes, with about 40% of phase one capacity earmarked for commercial peering. Microsoft, Google Cloud, and AWS have all signed letters of intent for India in country capacity inside the Yotta footprint, though firm contracts have not been disclosed.
How does this compare to Singapore's recent AI campus announcements?
Singapore's announced 2026 AI builds are smaller in raw GPU count but tend to ship sooner and at higher utilisation. Yotta's advantage is scale and Indian sovereign access; Singapore's advantage is speed and regional connectivity.
Comments
Sign in to join the conversation. Be civil, be specific, link your sources.