AirTrunk’s Entry into India: Decoding the Hyperscaler Land Grab and the Rise of AI Infrastructure as an Asset Class

Introduction: A Deal Richer Than It Appears – AirTrunk’s Strategic India Bet

On February 2025, AirTrunk—backed by Macquarie Asset Management and PSP Investments—announced the acquisition of Lumina CloudInfra, marking its formal entry into the Indian data center market. The transaction, valued at an estimated $700–900 million based on comparable recent deals in the Indian market, includes Lumina’s operational facilities in Mumbai and a significant land bank in Chennai and Hyderabad (Source 1: Industry deal disclosures).

This acquisition represents more than a geographic expansion. AirTrunk, which operates hyperscale campuses across Australia, Singapore, Japan, and Hong Kong, has identified India as the next frontier for AI training and inference workloads. The logic is structural: India’s data center capacity is projected to grow from approximately 1,200 MW in 2024 to 4,500+ MW by 2030, driven by hyperscaler demand (Source 2: CRISIL Infrastructure Report, Q4 2024).

The hidden axis of this deal is land and power grid capacity. AirTrunk is not merely acquiring colocation space. The Lumina portfolio includes 50+ acres of land parcels with assured power connectivity from state electricity boards, a constraint that has become the primary bottleneck for new data center construction in India. Similar land-grab dynamics have been observed with Brookfield’s 2024 acquisition of a 200-acre plot in Navi Mumbai and NTT’s expansion of its Chennai campus to 150 MW capacity (Source 3: JLL Data Center Market Report, January 2025).

India’s submarine cable infrastructure further amplifies the strategic value. The country is currently connected to 14 submarine cable systems, with three additional cables planned through 2027, positioning it as a critical node between Europe, the Middle East, and Southeast Asia (Source 4: TeleGeography Cable Database).

---

The Economic Logic: Why AI Data Centers Are Different (And Why India Wins)

Power Density and Architecture Divergence

AI data centers fundamentally differ from conventional cloud infrastructure. Traditional enterprise workloads operate at 5–10 kW per rack. AI training workloads, particularly those utilizing NVIDIA H100 or B200 GPU clusters, require 30–50 kW per rack, with some liquid-cooled configurations exceeding 100 kW per rack (Source 5: Uptime Institute AI Infrastructure Report, 2024).

This density shift has three implications for facility economics. First, it increases the per-square-meter revenue potential of data center space by 3–5x, fundamentally altering the underwriting metrics. Second, it requires purpose-built electrical and cooling infrastructure, including direct-to-chip liquid cooling and 220kV substation connectivity. Third, it concentrates power demand in single campuses, making load scheduling and grid interconnection the primary operational constraint.

India’s Competitive Power Position

India’s average industrial electricity tariff stands at approximately $0.08–0.10/kWh, compared to $0.12–0.15/kWh in Singapore and $0.10–0.12/kWh in Malaysia (Source 6: International Energy Agency, India Energy Outlook 2024). More critically, India has achieved solar tariffs as low as INR 2.60/kWh ($0.031) under competitive bidding, and round-the-clock renewable power PPAs have reached INR 3.50–4.00/kWh ($0.042–0.048) for hybrid solar-wind configurations (Source 7: SECI tariff data, Q3 2024).

These economics enable data center operators to lock in 15–20 year power contracts at rates 30–40% below Southeast Asian benchmarks, a structural cost advantage that directly impacts EBITDA margins. AirTrunk’s acquisition of Lumina, which has secured 250 MW of power connectivity approvals, effectively capitalizes on this arbitrage.

Tax and Regulatory Environment

The Indian Goods and Services Tax (GST) regime applies 18% GST to cloud services but provides for input tax credit on data center infrastructure services. This creates a vertical integration incentive: operators that own both the infrastructure and provide colocation services can optimize their tax position. Additionally, state-level policies in Maharashtra, Telangana, and Tamil Nadu offer 100% electricity duty exemptions for 5–10 years and expedited environmental clearances for data center projects exceeding 50 MW (Source 8: India Data Center Policy Tracker, ICRA, 2024).

Fiber Latency and India as an AI Inference Hub

India’s geographic position provides sub-100ms latency to both European and Asia-Pacific markets via existing cable systems. The planned SEA-ME-WE 6 and India-X cables will further reduce latency for AI inference workloads that require real-time processing between user bases in the Middle East, Africa, and Southeast Asia (Source 9: Submarine Cable Network Annual Report, 2024). This positions Indian data centers as viable nodes for serving inference traffic from a 2.8 billion person user catchment area.

---

Who’s Really Behind the Curtain? Blackstone, SoftBank, and the Pension Fund Playbook

The Investor Syndicate Dynamics

The AirTrunk ownership structure—Macquarie Asset Management (majority owner) and PSP Investments (Canadian public pension fund)—represents a long-duration, yield-oriented capital base. Macquarie’s infrastructure fund typically targets 12–15% IRR with 20+ year holds, while PSP seeks inflation-hedged assets with stable cash flows (Source 10: Macquarie Infrastructure Investor Conference, 2024). This contrasts with Blackstone’s opportunistic approach, which has deployed $6 billion into Indian data center assets since 2023 through its portfolio company BAM Digital Realty.

The divergence matters for competitive dynamics. AirTrunk’s capital structure allows for patient land banking and pre-leasing strategies. Blackstone, by contrast, has accelerated development timelines, constructing speculatively and leasing post-completion. SoftBank’s interest in Indian AI infrastructure, as reported through Vision Fund discussions with multiple operators, signals a third approach: aligning data center investments with Arm-based chip deployments and edge AI workloads (Source 11: Financial Times, SoftBank India AI Strategy, January 2025).

Asset Class Valuation Mechanics

Hyperscale data centers are now traded on comparable multiples to tower infrastructure and fiber assets. Recent transactions in India and Southeast Asia have closed at 20–25x trailing EBITDA, with forward multiples compressing to 18–22x as new supply enters the market (Source 12: CBRE Asia Pacific Data Center Capital Markets Report, Q4 2024). For AirTrunk’s Lumina acquisition, assuming a 25x EBITDA multiple on stabilized operations, the deal implies annual EBITDA of $28–36 million—indicating significant development upside given Lumina’s current 35% occupancy rates.

Risk Assessment

The primary risk factors for AirTrunk’s India strategy include:

• 5G rollout dependency: AI inference workloads require 5G network slicing capabilities for low-latency edge deployment. India’s 5G population coverage reached 40% in Q4 2024, with full metropolitan coverage expected by 2026 (Source 13: TRAI Connectivity Report). Delays would suppress demand for edge AI services.
• Land acquisition timelines: Indian data center projects currently require 18–24 months for land title verification, environmental clearance, and grid connectivity approval. State-level differences in clearance speed—Tamil Nadu averages 12 months versus 20 months in Maharashtra—create execution variance (Source 14: NASSCOM Data Center Infrastructure Report).
• Power grid reliability: Despite improvements, India’s grid experienced 5.2 hours of average annual outage per customer in 2024, requiring operators to maintain 2N+1 backup diesel or battery systems, which adds 15–20% to capital expenditure (Source 15: Ministry of Power, Annual Grid Performance Data).

---

Deep Dive: Lumina CloudInfra’s Hidden Value – Land, Power, and Fiber Constraints

The Lumina CloudInfra portfolio provides a case study in the three binding constraints that will determine asset value over the next decade.

Land Constraints

Available industrial land within 50km of India’s three primary fiber hubs—Mumbai, Chennai, and Hyderabad—has declined by 60% since 2020, driven by competing demand from manufacturing (PLI scheme), logistics parks, and residential development (Source 16: Knight Frank India Industrial Land Report 2024). Lumina’s land parcels in Ambegaon (Mumbai) and Maheshwaram (Hyderabad) were acquired between 2021 and 2023 at prices of INR 15–25 crore per acre, representing a 2–3x appreciation in three years based on comparable transactions.

Power Constraints

India’s peak power deficit, though reduced to 0.4% nationally, remains 3–5% in data center heavy states like Maharashtra during summer months. Data center operators must secure “firm power” status from state load dispatch centers, a process that Lumina had completed for its Chennai campus. This approval is a transferable asset that accelerates AirTrunk’s construction timeline by 8–12 months relative to greenfield competitors (Source 17: POSOCO Grid Connectivity Database, November 2024).

Fiber Backhaul

The Mumbai-Pune-Nashik fiber corridor, which hosts 85% of India’s international internet bandwidth, is operating at 70% utilization, with planned expansions from Bharti Airtel and Reliance Jio not expected until 2026 (Source 18: India Internet Exchange Traffic Report 2024). Lumina’s pre-existing dark fiber lease agreements with these providers provide guaranteed 100 Gbps connectivity with 48-hour repair SLAs, a contractual position that new entrants cannot easily replicate.

---

Market Predictions and Neutral Assessment

The AirTrunk-Lumina acquisition signals a structural shift in how global capital allocates to Indian digital infrastructure. Three predictions emerge from this analysis:

• The asset class will bifurcate: By 2027, Indian data centers will split into two distinct categories—AI-optimized facilities (power density >30 kW/rack, liquid cooling ready) commanding 25–30x EBITDA multiples, and traditional colocation facilities trading at 15–18x. AirTrunk’s portfolio is positioned in the former category.
• Land and power will outperform operational earnings: The appreciating value of approved land parcels and grid connectivity permits will constitute 40–50% of total portfolio returns over the next five years, based on observed appreciation in Singapore and Tokyo markets during similar expansion cycles (Source 19: Data Center Dynamics, Asset Value Analysis 2023–2024).
• Consolidation will accelerate: With total institutional capital targeting Indian data centers estimated at $25–30 billion through 2030, the current market of 150+ operational operators will consolidate to 15–20 scaled platforms. AirTrunk, with Macquarie’s backing, possesses the capital capacity to acquire and develop 500+ MW of capacity, positioning it among the top five operators.

The transaction does not represent a thesis on Indian GDP growth or digital adoption—those are premises already priced into valuations. It represents a bet on the irreversibility of AI infrastructure concentration and India’s capacity to serve that demand within the constraints of its physical and regulatory infrastructure. The final determinant of returns will not be cloud revenue but the ability to navigate land acquisition and power procurement, competencies far removed from traditional data center operations.