Asia’s data centre market is in the midst of an infrastructure supercycle. In 2025, Asia-Pacific data centre investment reached a record US$11.6 billion, with Malaysia, Australia, and India emerging as the new centres of gravity for AI-driven capacity growth. Yet this expansion is colliding with hard constraints: power availability, fragmented data sovereignty regimes, and a persistent skills gap.
For CIOs, the challenge is no longer simply procuring compute—it is architecting infrastructure that is simultaneously AI-ready, compliant across divergent jurisdictions, and sustainable enough to satisfy regulators, investors, and customers alike.
Govind Choudhary, GM of Southeast Asia and India at Digital Realty, argues that the industry must move beyond passive sustainability measurement. “There’s already a lot of data that is being collected,” he says. “The priority now needs to be how actually to use this data better.”
His prescription is operational: embed real-time sustainability metrics into cooling, power, and water management decisions, and design new capacity with these metrics upfront rather than reporting them retrospectively.
The power-geography reset
CBRE’s 2026 Asia Pacific Data Centre Trends & Outlook Report describes a market undergoing a “significant reordering,” with growth shifting from traditional Tier I hubs toward power-advantaged locations.
Johor in Malaysia recorded a 53% year-on-year rise in live capacity in 2025, making it the region’s fastest-growing market by that measure, while Melbourne grew 37%. Singapore and Hong Kong, by contrast, grew at roughly 6% to 8%, constrained not by demand but by power and land limitations.
This geographic redistribution has a direct implication for CIOs: location strategy is now inseparable from energy strategy. Choudhary emphasises that “power availability is a key requirement,” and that CIOs must “choose locations where there is good long-term availability of power, it’s cost-effective, the grid is reliable, and there’s availability of renewable energy.” Crucially, he frames this as “a continuous evaluation process, not a one-time decision.”
The scale of the power challenge is stark. Single AI training clusters can consume 30–100 MW, equivalent to a medium-sized traditional data centre. Southeast Asia now accounts for approximately 50% of all data centre capacity under construction in APAC, with Malaysia leading at 1,039 MW under construction and Thailand following at 859 MW.
Andrew Green, head of Data Centre Group, Asia Pacific at Cushman & Wakefield, notes that “the challenge increasingly centres on securing the power infrastructure needed to support the next generation of AI workloads”.
Data sovereignty as an architectural constraint
While power reshapes geography, data sovereignty reshapes architecture. Across APAC, regulators are tightening rules on cross-border data movement. Singapore’s PDPA Transfer Limitation Obligation requires “comparable protection” for overseas data transfers.
Australia’s APP 8 obligates organisations to ensure offshore recipients do not breach Australian Privacy Principles. South Korea’s Personal Information Protection Act (PIPA) demands detailed consent for cross-border transfers, including destination, purpose, and retention period.
Vietnam’s new Law on Personal Data Protection, effective January 2026, introduces strict consent criteria and data protection rules for financial services, metaverse, and big data activities. South Korea’s AI Basic Act, which took effect in January 2026, is the first comprehensive AI law in Asia, imposing stringent oversight on “high-impact” AI applications and potentially forcing a “highest common denominator” approach across the region.
Choudhary does not see sovereignty and scale as opposing choices. “We spoke about the federated model and how it can actually help CIOs really manage both the requirements,” he says, “ensuring that they can place the sensitive, regulated data in the country where they need to be hosted locally to meet the local compliance requirements.”
The key enabler is connectivity:
“If they create a distributed architecture that is well connected and secure, then what they can do is they can really continue to meet the sovereignty requirements by placing sensitive data in-country and connect it back to the regional architecture where they have centralised resources for compute.” Govind Choudhary
This federated approach is echoed in industry research. Akamai’s 2026 Cloud and Security Predictions for APAC anticipates that “digital sovereignty becomes economic sovereignty,” with organisations viewing cloud portability as essential risk mitigation against geopolitical uncertainty.
“AI is fundamentally changing the economics of cyberattacks in APAC. Adversaries are no longer scaling through workforce, but rather through automation. Leaders can’t rely on human-paced defences in a machine-paced threat environment,” said Reuben Koh, director of Security Technology & Strategy at Akamai.
The same report predicts that 80% of APAC CIOs will rely on edge services for AI performance and compliance by 2027.
Designing for AI: Not all workloads are equal
A central theme in Choudhary’s guidance is workload differentiation. “Are they designing for training? Is it for fine-tuning of models or is it for inference?” he asks.
The answer determines architecture. Training workloads are “mostly power intensive” and “don’t need to be close to end users so that they can be centralised.” Inference, by contrast, “needs to sit closer to end users and data,” demanding “a more distributed model.”
This distinction maps onto a broader industry shift toward distributed AI architectures. Akamai predicts “stronger momentum behind distributed AI architectures, as enterprises move inference closer to users and operational systems to improve latency and performance”.
Choudhary also cautions against assuming that all AI workloads belong in the public cloud. “Public cloud is great,” he acknowledges. “It provides agility, and it provides a way to scale up fairly quickly. However, as you scale up, the cost can also be quite prohibitive.”
The implication is a hybrid strategy that balances public cloud agility with private colocation control for cost, cybersecurity, and data privacy.
Connectivity becomes the connective tissue of this distributed model. “Through good connectivity, what they can do is distribute the compute resources where it’s needed and not necessarily have to deploy compute resources in every region,” Choudhary says.
He says an infrastructure connectivity layer, like Digital Realty’s Service Fabric, is positioned as the mechanism for this: “It allows organisations to connect all the datasets across the world and seamlessly balance workloads and manage the infrastructure.”
The talent bottleneck
A successful infrastructure strategy, Choudhary insists, “will only work if you have enough competent people with the right skill sets to really design, run, and operate it.” He frames talent development as “an important part of infrastructure roadmap” rather than “just an HR issue.”
The skills required—liquid cooling, high-density power management, sustainability reporting, critical infrastructure maintenance—are “not something that is off-the-shelf skills easy to acquire from a university.”
Digital Realty’s response includes internships and graduate programmes, partnerships with universities for curriculum input and guest lecturing, and certification pathways developed with third-party academies such as DCD Academy.
Choudhary also advocates for ecosystem-wide collaboration: “Rather than just necessarily competing with each other, if operators work with their customers and partners to create joint programs, which also allows talent to rotate across different… it can give them unique skill sets that can really help boost the talent shortage in the industry.”
This skills gap is corroborated by Alibaba Cloud’s 2026 survey of Asian enterprises, which found that 37% of respondents identified insufficient internal expertise as a major barrier to AI adoption, with 49% calling for better talent and skills development. Data privacy and security concerns ranked as the top obstacle at 48%, followed by implementation costs at 42%.
Sustainability as competitive differentiator
Choudhary argues that sustainability, properly embedded, creates value beyond compliance. “Organisations that have clear sustainability goals can have good favour with regulators,” he says. “It can create regulatory advantage, potentially faster approvals and access in constrained markets.”
Second, it attracts investment: “Many long-term capital providers have sustainability goals of their own, and they’d like to work with organisations that have a clear path on achieving those goals.” Third, it attracts talent, particularly younger employees who evaluate organisations on their sustainability policies and practices.
The cost argument is nuanced. “Even in the short term, some of that cost may be higher. The overall longer-term cost will be lower.” This reframing positions sustainability not as a premium but as a long-horizon efficiency play—one that becomes a “competitive advantage for organisations that really design this bottom-up thinking about the long term.”
Selecting the right partner
For CIOs navigating this complexity, Choudhary offers a checklist for operator selection: clear sustainability goals, regional and global coverage rather than a single location, connectivity infrastructure, and long-term credibility with utility providers.
“When organisations choose infrastructure partners, they’re really planning for the long term,” he says. “They really need to ensure that the partners, the operators they have chosen, have the credibility, track record, and ability to sustain the services in the longer term.”
The federated model he describes—sensitive data in-country, compute distributed across regions, connected through a secure fabric—offers a template for balancing sovereignty with scale.
Whether CIOs can execute it depends on whether they treat infrastructure as a strategic differentiator or a procurement exercise. Choudhary’s message is unambiguous: the former is now the only viable option.
Click the PodChats player to hear Choudhary’s take on key data centre trends in Asia, shaped by accelerating AI adoption, evolving regulation, and the challenges of sustainably supporting a business in the AI era.
- How do you see data centres in Asia addressing the growing gap between sustainability data collection and meaningful action?
- How should CIOs, in partnership with data centre operators, navigate the diverging regulatory landscapes across the region?
- Given the accelerating use of AI, how should CIOs design and build the right infrastructure for AI workloads while maintaining sustainability?
- How should CIOs, again working with DC operators, address the skilled workforce requirements to execute this strategy?
- How should CIOs manage sovereign data requirements while leveraging regional scale?
- Climate change, volatile fuel costs, the accelerating use of resource-hungry technologies like AI, – all these pose serious challenges to keep data centres operating at peak performance. Any recommendation on how to build resiliency against power constraints and grid instability?
- On the topic of sustainability, how should CIOs embed sustainability as a competitive differentiator, not just a compliance cost?
- Not all DC operators are equal. What is your advice when selecting their operator/partner to address calls for embedding sustainability not just in their AI infrastructure but across all their compute needs – local and international?
