The rapid growth in demand for artificial intelligence (AI) has made the physical infrastructure underpinning it an increasingly material enterprise risk, forcing organisations to strictly manage the energy, water, land and community impacts of AI datacenters.
Key Insights
- The rapid build-out of high-density AI datacenters threatens to outpace municipal resource capacity, making infrastructure governance a critical pillar of enterprise risk management rather than a mere compliance checklist.
- Operators should improve both energy efficiency and power density, including consolidating inefficient legacy facilities where appropriate. Singtel’s transition of its legacy facilities to a single AI-ready facility illustrates how newer facilities can use substantially less land while delivering greater capacity translating to a competitive operational advantage.
- Regulators across the Asia-Pacific region are increasingly using strict gating mechanisms for new datacenter development. Applicants may need to demonstrate not only efficient operations and credible green energy pathways, but also wider benefits to the infrastructure, economy and communities of the markets in which they operate to secure approval.
The rapid acceleration of AI requires infrastructure that can place significant pressure on global energy grids, water supplies, and land use. When evaluating AI and sustainability, organisations must therefore look beyond software efficiency to consider the infrastructure that makes these technologies possible and their respective impacts.
Scaling these capabilities sustainably without compromising long-term enterprise value requires business leaders to abandon a ‘growth at all costs’ mindset. Risk practitioners need to actively manage resource constraints, treating infrastructure governance as a core component of corporate strategy rather than a simple compliance check.
Speaking at the Institute of Enterprise Risk Practitioners’ (IERP®) Global Conference 2026, Ivan Li, Founder and Managing Director of Climant Impact, highlighted this stark reality. He warned that the exponential demand for compute power has transformed AI infrastructure into a direct operational and environmental risk. However, this would also create opportunities to strengthen the systems on which datacenters depend.
The Rise of AI Infrastructure
Putting this exponential surge into perspective, SemiAnalysis’ projections indicate that global datacenter capacity will double by 2028, with over half of this new build-out dedicated specifically to AI infrastructure. This indicates a rapid shift towards higher-density infrastructure required for AI workloads.
“You’re no longer looking at small-level datacenters of five to ten megawatts,” Ivan explained. “You’re looking at 100, 200 megawatt campuses, gigawatt campuses.”
If left unmanaged, this physical build-out can strain shared infrastructure and natural resources, erode community support and ultimately constrain the stable growth of datacenter capacity.
Enterprise Risk Management: Measuring the Hidden Resource Strain
Evaluating these physical constraints now sits squarely within the discipline of enterprise risk management. As mega-campuses outpace local grid build cycles, companies must account for three major risks:
- Energy and Grid Capacity: Datacenter can compete with other users for limited firm power and grid capacity, potentially contributing to higher cost and local constraints unless new clean supply and network capacity are developed alongside demand.
- Water Scarcity: Higher-density AI workloads increase cooling requirements. The appropriate choice among air, hybrid, and liquid cooling depends on the local water-energy trade-off; where potable water is used in water-stressed markets, it places these facilities in direct competition with local municipal needs.
- Embodied Carbon: The construction of gigawatt campuses requires massive amounts of steel, concrete and equipment, generating heavy upfront emissions. Since these emissions are largely determined during the design and procurement phase and cannot be reduced retrospectively through efficient operations, making early supplier and material choices are critical.
Summarising the sheer scale of this footprint, Ivan illustrated: “When you look at a typical real estate commercial building and compare it to a datacenter, it could be the same amount of space, but it could be 100 times in terms of the energy consumption.”
Assessing these resource demands helps business leaders understand the wider environmental and community implications of their digital ambitions.
Optimising the Material Footprint of Gigawatt Campuses
As these mega-campuses expand, land becomes a contested asset and older facilities simply lack the processing power for modern AI workloads. Every hectare claimed for a datacenter competes with industrial and residential needs, compelling operators to improve power density rather than sprawl outward.
Singtel’s recent consolidation in Singapore exemplifies this shift. The provider decommissioned five legacy datacenters and consolidated the load into a single, AI-ready facility. This next-generation site occupies less than one-tenth of the footprint of the pre-existing facilities while reducing design PUE from more than 2.0 to 1.25, an improvement of approximately 43%.
Acknowledging the commercial realities of upgrading AI infrastructure, Ivan framed consolidation as a strategic necessity.
Ivan acknowledged that consolidating legacy facilities requires significant upfront investment and is not a simple process. However, moving to a newer, more efficient facility can improve resource efficiency while releasing scarce land for other uses.
Reducing both land and energy intensity allows operators to transform a logistical constraint into a competitive operational advantage. Yet, as scarcity pressures mount, governments are no longer leaving these optimisations to voluntary corporate goodwill.
Evolving Governance and the Push for Green Datacenters
Technology alone cannot resolve the full range of resource and community impacts. Governance frameworks increasingly mandate that the development of green datacenters follow a model of responsible and sustainable expansion.
Regulators across the Asia-Pacific region are shifting away from permissive growth in favour of conditional access, enforcing strict requirements for new builds. Industry expectations are also rapidly converging on firm baselines, including a push for 100% renewable energy utilisation by 2030 and a Water Usage Effectiveness of 2.0 or lower, although requirements vary by standard and local operating contexts.
This more selective approach is illustrated by reports that Johor recently rejected 30% of datacenter applications for falling short of sustainability standards. Singapore applies a particularly structured model through its Data Centre Call for Applications (DC-CFA) framework. Rather than solely reviewing capital investments, the framework evaluates applications based on broader environmental and economic contribution and benefits to the local datacenter ecosystem and beyond.
Hence, operators must demonstrate how their projects strengthen the systems and communities around them, for example by commercialising novel green energy pathways like biomethane, investing in international connectivity, creating high-value local jobs and supporting the wider datacenter ecosystem. Ivan outlined how these gating mechanisms can encourage system-level change rather than merely compliance with minimum thresholds.
“What they’re trying to do is force these datacenters to say, if you think this AI demand is real and you think that it creates that much value, you must have a corresponding investment from it from a sustainability perspective to drive towards the right outcomes,” he emphasised.
The Future of AI and Sustainability
The sustainability of AI growth depends on the industry’s ability to support it structurally, environmentally and economically. Forward-thinking organisations treat this reality not as a roadblock, but as a strategic advantage. By proactively addressing material constraints, operators turn a potential crisis of scarcity into a catalyst for innovation.
In the end, companies that optimise their AI infrastructure today will build more resilient business models, secure higher valuations, and lead the market tomorrow. As Ivan concluded in his closing remarks, “Managing these risks can be difficult, but it does also provide an opportunity to drive a positive impact.”
