With the structural development of artificial intelligence (AI) and stepped-up AI investments globally, the data center industry is undergoing a massive transformation. Following the Asia Communacopia + Technology Conference, we asked Goldman Sachs analysts Ronald Keung and Timothy Zhao about what may come next for data centers in Asia, the critical power bottlenecks, and how AI is reshaping the industry's landscape. Below are their shared responses:

What is the outlook for data centers in Asia?

The outlook for data centers in Asia is exceptionally robust, characterized by a broad and steep growth curve. Traditional cloud workloads continue to grow alongside new GPU and AI-first demand, expanding the total market size rather than simply redistributing it.

In China specifically, our team forecasts data center demand to grow at a 20% compound annual growth rate (CAGR) from 2025 to 2028, driven by structural developments in AI and stepped-up investments by hyperscalers and large language model (LLM) companies.

What are the biggest data center markets in Asia, with the most upside?

The SIJORI (Singapore–Johor–Riau Growth Triangle), established between Indonesia, Malaysia, and Singapore to strengthen economic links, has emerged as a strategic node in the region for the interconnected data center ecosystem over the past couple of years. Looking ahead, based on our recent Asia Communacopia + Technology Conference panel, our speakers see India, Japan, and the Philippines as key growth markets in the broader Asian region.

The panelists expect India to grow rapidly, supported by favorable demographics, deep engineering talent, and its strategic proximity to the Middle East. Japan was flagged as a standout market, with government-backed initiatives driving aggressive expansion. Finally, they see the Philippines emerging as a strong growth market due to reduced red tape and scalable power availability.

Could you say more about power being the key bottleneck in this sector in Asia?

Power remains a primary constraint affecting construction timelines and delivery capabilities across the region. Resolving grid infrastructure issues typically spans years to decades, making it a significant hurdle. Meanwhile, in China, domestic chip production is the near-term constraint, while overall power supply is much less of a constraint for data center development compared to the rest of Asia, as data centers have only accounted for around 2% of China’s power consumption of late. However, local green power supply and strict government requirements—such as mandating that green power account for over 80% of new computing infrastructure in major hubs—can act as short-term bottlenecks.

How can the power bottleneck be resolved to enable the growth of this sector?

Public-private partnerships are essential to bridging the infrastructure gap. Governments should take the lead given that power is often a state-owned or heavily regulated utility, while private operators and hyperscalers are increasingly investing directly in power generation and renewables. Malaysia was highlighted at our conference as a prime example where government-utility collaboration continues to progress.

Additionally, operators are adopting tailored approaches to balance power usage effectiveness (PUE) and water scarcity. For instance, some facilities in Mumbai are sourcing treated wastewater from nearby purification plants rather than using potable water, achieving highly efficient PUE levels.

How are shifting customer mixes affecting data center operators?

The rise of AI is fundamentally changing customer timelines and infrastructure needs. Single-tenant AI customers are committing to orders well before construction starts to lock in bespoke designs, whereas traditional multi-tenant customers expect much shorter lead times.

Furthermore, new AI-first customers and neoclouds move at a radically different commercial speed, pushing for modular and prefabricated designs to compress delivery timelines; new contracts can close in just six weeks compared to four to six months for traditional hyperscalers.

Speakers at our recent Asia Communacopia + Technology Conference noted that operators in Southeast Asia are preserving fungibility by making hybrid air and liquid cooling the default design choice and engineering facilities to support highly variable rack densities, ensuring they can cater to both GPU-intensive AI and traditional cloud workloads. In China, evolving customer demand is enabling a geographic shift towards power-rich Western China hubs as the new markets for capacity expansion, along with new business models for data center operators such as AI GPU infrastructure rental.

What does the regulatory landscape for data centers look like across Asia?

Geopolitics and regulatory shifts have introduced higher compliance overhead, resulting in operational adjustments, such as enhanced due diligence that now extends beyond direct customers to ultimate end-users and supply chain participants. However, these shifts have not deterred the underlying demand momentum.

Coordination between computing power and electrical power in terms of both quantity and dispatch serves as the core regulatory intelligence. For example, Malaysia has announced power tariffs for data centers and limits on non-AI related data center development over the past year.

In China, the regulatory landscape is heavily shaped by national strategies like the "East Data, West Computing" project, which promotes concentrating computing facilities in western regions rich in green power, and more recently by window guidance on new data center approvals to control capacity supply growth and align it with national computing hubs.

This article is being provided for educational purposes only. The information contained in this article does not constitute a recommendation from any Goldman Sachs entity to the recipient, and Goldman Sachs is not providing any financial, economic, legal, investment, accounting, or tax advice through this article or to its recipient. Neither Goldman Sachs nor any of its affiliates makes any representation or warranty, express or implied, as to the accuracy or completeness of the statements or any information contained in this article and any liability therefore (including in respect of direct, indirect, or consequential loss or damage) is expressly disclaimed.