Singapore Telecom (STEL.SI) and ASEAN Utilities: Primary Beneficiaries of the AI Infrastructure Supercycle

We are at a structural inflection point driven by AI infrastructure build-out. ASEAN, with its strategic location, available resources, and supportive policies, is a core beneficiary. Telecom operators, led by Singtel, and utility companies will gain multi-year growth from the rapid expansion of data center capacity and the associated premium for reliable power. Market forecasts likely understate the scale of ASEAN’s build-out and the structural demand shift from Agentic AI.

Global Capex Tsunami Provides a Decade-Long Tailwind

The commitment from US hyperscalers ensures sustained demand for ASEAN data center capacity. The five largest US cloud/AI providers have committed to nearly $690 billion in capital expenditures for 2026 alone, almost double 2025 levels. Hyperscale capex is projected to grow at a ~45% CAGR through FY28e, approaching $1 trillion. This has driven an upward revision of ASEAN AI infrastructure capacity forecasts to 19GW by 2035e, implying a 19% CAGR and a 6x expansion from 2025 levels. This capital cycle provides visibility and underpins multi-year growth for infrastructure providers in the region.

Agentic AI Multiplies Inference Demand

The new paradigm of Agentic AI, exemplified by frameworks like OpenClaw, structurally increases computing demand. The shift from a ‘pull’ model (user queries) to a ‘push’ model (agents operating autonomously 24/7) will multiplicatively increase demand for computing infrastructure, particularly for inference. Technological advancements like NVIDIA’s Vera Rubin platform, which promises up to a 10x reduction in inference token cost, will accelerate adoption. This evolution ensures that the current infrastructure build-out is not a one-time event but the foundation for persistent, growing demand.

Singtel is the Regional AI Infrastructure Leader

Singtel’s Digital InfraCo (Nxera) is the clear regional leader, positioned for accelerated growth and re-rating. Its total capacity is set to surge from 62MW in FY25 to >200MW in FY26e, with a medium-term target exceeding 400MW. Its partnership with NVIDIA to build an AI Centre of Excellence and its GPU-as-a-Service offering via the Paragon platform creates a full-stack ecosystem. The recent acquisition of a 25% stake in STT GDC provides instant access to a 1.7GW pipeline, scaling its presence to 12 markets. We forecast the Digital InfraCo segment to contribute ~15% of group EV by FY28e, valued at a 20x EV/EBITDA multiple versus 5-6x for core telco operations.

Power is the Ultimate Bottleneck, Benefiting Utilities

Data center growth is ultimately a power story, creating a premium for reliable electricity. Next-gen GPU platforms will require 600kW-1MW per rack, 3-5x current deployments. ASEAN’s data center power demand could reach ~16.6GW by 2030, representing ~12% of regional generation capacity, up from ~5GW in 2025. In Singapore, data centers already consume ~25% of national electricity. This tightens power markets, benefiting utilities and independent power producers (IPPs) like Keppel, Tenaga Nasional, and Gulf Energy Development through higher pricing power, new growth avenues, and expanded regulated asset bases.

Regional Growth Paths Diverge

ASEAN data center growth is robust but uneven, favoring markets with policy support and power availability. Thailand’s capacity forecast has been upsized to 3GW by 2035e (from 1GW), a ~28% CAGR, backed by >$3.1bn in newly approved projects. Malaysia, with low-cost land/power and proximity to Singapore, has attracted massive investments (e.g., Amazon’s $6.2bn commitment), with potential demand of ~12GW. Indonesia’s growth (~13% CAGR) is constrained by grid reliability, though Batam benefits from Singaporean spillover. The Philippines shows the fastest percentage growth (~37% CAGR) from a small base.

Key Risks

The primary risk is a slowdown in the hyperscaler capex cycle, which would directly dampen data center demand. Physical and regulatory bottlenecks—power grid constraints, land/water scarcity, and permitting delays—could push out project timelines. The diffusion of Agentic AI to ASEAN may take longer than the expected 1-2 years. Lastly, regulatory frameworks in markets like Malaysia could tighten, affecting project economics.

Valuation and Trade Implications

For Singtel, the re-rating story hinges on the high-growth Digital InfraCo segment. Applying a 20x EV/EBITDA multiple to this business supports a Sum-of-the-Parts valuation, with a price target of S$5.50. For utilities, the investment case is driven by rising, inelastic power demand, which supports earnings upgrades and potentially higher valuation multiples. Investors should focus on companies with visible data center-linked capacity expansion and strong partnerships.

Appendix Data Summary

ASEAN Data Center Capacity Forecast (GW)

Selected ASEAN Telco DC Projects

• Singtel (STEL.SI): DC Tuas (58MW) operational; JV in Johor (64-200MW); JV in Batam (51MW); GSA JV in Thailand (26MW, >80% pre-sold).
• AIS (ADVANC.BK): GSA01 (25MW) operational; GSA02 (38MW) for 1H27.
• PLDT (TEL.PS): Building 12th DC (100MW); long-term target 500MW.