Asia-Pacific data centers are swinging into investors' view. We estimate more than US$100 billion will be invested in such facilities in the region over the next five years. The spending will capitalize on strong data growth and the rise in AI, cloud computing and digitalization.

Rated telecom companies in the region are also increasingly investing in, and operating, data centers. They will be the growth engine for many telecoms and real estate companies.

According to estimates from S&P Global Market Intelligence 451 Research, data center capacity in Asia-Pacific will likely grow 17% annually (compounded) through to 2029. The recent US$16 billion acquisition of Asia-Pacific based operator AirTrunk by Blackstone Inc. and Canada Pension Plan Investment Board, underscores the importance of the Asia-Pacific market.

Through this boom, operators will need to manage operating risks, cost volatility, obsolescence, resource constraints and evolving regulations. As such, investors are asking us about the risks of investing in this relatively new asset class. From a corporate rating perspective, we address their queries below.

Frequently Asked Questions

What are the main trends for data centers in Asia-Pacific?

In a nutshell, accelerated expansion. We expect emerging markets in Asia-Pacific to outpace established markets in capacity growth over the next three to five years (see table 1).

Key emerging markets include India, Indonesia, Malaysia, Philippines, Thailand and Vietnam. Key established markets include Australia, China, Hong Kong, Japan, Singapore and South Korea.

Emerging Asia-Pacific countries are increasingly an attractive alternative to developed markets for three reasons:

• Lower costs in emerging Asia-Pacific to develop and operate data centers;
• Significant growth in data demand due to digitalization and favorable demographics within emerging Asia-Pacific; and
• Government support for the building of local data centers as part of their push for data sovereignty.

Table 1

Johor Bahru, in Malaysia, for example, has become a hot landing spot for new data centers. The Johor Bahru sites can be a cheaper alternative to Singapore for regional data traffic, with more affordable land and power. The government is supporting the data center developments by building power and connectivity infrastructure.

An increasing focus on data sovereignty (see table 5 in the Appendix) has spurred the building of onshore data centers to store government data and sensitive personal information. This will also result in less reliance on foreign-based data center operators.

For example, the Australia government has moved some sensitive data out of data centers operated by Global Switch Holdings Ltd., after a China-backed consortium bought a 49% stake in the company.

Are the business risks lower for data center operators focused on established Asia-Pacific markets?

Yes, in general. Operators with data centers mainly in established markets face lower operational risk than equivalent facilities in emerging Asia-Pacific. The former benefit from lower country risks, which translate into a relatively stable policy environment, a well-established power and connectivity infrastructure, and a more digitalized economy.

Data centers in tier-one cities (Singapore, Tokyo and Sydney, etc.) will likely maintain low vacancy rates and higher pricing power. This is due to their strategic location close to end users that are generating and consuming data. This adjacency factor will be critical for workload requiring low latency and seamless connectivity. Scarce land and power resources in key cities will keep supply limited.

Chart 1

Workload with lower latency requirements in established markets will be pushed to tier-two or tier-three cities, easing resource constraints. For example, Japan is offering subsidies to decentralize data-center development, such that new centers will be built away from Tokyo and Osaka. Data centers in tier-two or tier-three cities typically have lower land and construction costs compared with tier-one cities.

The competitive landscape in emerging markets is less predictable given the fast pace of development in those regions, and the influx of international players. For example, Johor Bahru has a project pipeline of 1.67 gigawatts (GW) of total IT load capacity, which is markedly higher than its current capacity of 231 megawatts (MW) as of the first half of 2024, according to property broker Cushman and Wakefield. We see a risk of transitional oversupply when such capacity enters the market.

How do the revenue models of data center operators affect the quality of their earnings?

The characteristics of data center earnings depend heavily on whether there is a wholesale or retail focus. The two models differ in terms of the customer mix, contractual terms, service offerings and associated risks.

Table 2

The hyperscale or wholesale colocation model generally offers operators a high predictability of earnings and cash flow. Our analysis views such features as credit positive.

This model, combined with a high ownership of assets, may support a higher leverage tolerance at a given rating level. A high ownership of assets may help an operator eliminate the risk surrounding renewal of property and land leases, and offer greater financial flexibility for fundraising. Digital Realty Trust Inc. (BBB/Stable/--), a U.S.-based data center operator, has such qualities, for example.

This is because wholesale contracts are long, with customers committing to a fixed payment or a minimum payment even if their actual usage falls short. There is also a high penalty for prematurely terminating contracts.

The high switching costs could reduce renewal risks. This is because, in making the move their customers would need to replicate or relocate vast amounts of physical hardware, network systems and storage infrastructure.

For the retail colocation model, the renewal risks and cash flow predictability could vary depending on factors such as the level of interconnectivity and--to a smaller extent--the existence of bundled services. These interconnections, which aim to create network effects by building direct, secured connections among customers, are hard to replace. They also attract customers that would like to establish connections within an operator's network.

Large retail colocation providers or operators with significant retail colocation exposure are creating a stickier customer base with interconnection. Examples include Equinix Inc. (BBB/Stable/--), a U.S. based data center operator. Smaller retail colocation providers may not be able to offer a competitive interconnection services.

The retail colocation model requires less investment and lower levels of customization. Otherwise, retail contracts are shorter, exposing operators to shifts in demand and market volatility. They also offer services that have lower earnings predictability than the rental of space and power. This may include remote technician services, network monitoring and cloud services.

Most of the retail colocation centers are speculative builds without sizable pre-committed capacity. If operators can't ramp up their utilization, this will hurt margins and cash flow.

What risks do wholesale operators face given their high tenant concentration?

Wholesale operators' data centers are often customized to fit customer requirements. The value of operators' assets would fall if their key customer faced financial distress; it could be costly to reconfigure the data center to match a new customer, should the old one drop out.

The strong credit qualities of customers can substantially mitigate counterparty risk. Most clients of wholesale operators are cloud providers or large internet companies. The entities are cash generative and would likely be able to honor their lease contracts. The contract termination clause will add another layer of protection for early termination.

There are other risks. For example, customers have significant bargaining power regarding pricing. It's common to see a few large customers, usually hyperscalers or large internet/media companies, taking up most of a facility or even an operator's total capacity. In many cases, these important customers are also offtakers for new projects and when expanding into new regions.

Such bargaining power is apparent in Asia-Pacific markets that have lower utilization rates and abundant new supply, such as China. This may squeeze operators' margins both on new developments and during renewals.

Some hyperscalers are also building their own data centers in strategic locations. This could increase renewal risks for their service contracts with operators. This so-called insourcing gives hyperscalers more control over their data center infrastructure. For instance, Meta Platforms Inc. completed its data center in Singapore in 2022.

Hyperscalers will likely weigh such benefits against other considerations, such as the pace of expansion and local knowledge.

How well can data center operators pass through costs and protect their margins?

In our view, wholesale operators generally have better cost pass-through ability. This is through terms that are embedded in their lease contracts that otherwise expose them to more cost fluctuations over the long duration of the contract. We believe such arrangements add resilience to profits and cash flow.

Wholesale/hyperscale operators typically use so-called triple-net and modified gross leases, which require the tenants to reimburse the manager for electricity and other costs. Under both types of leases, tenants are responsible for all costs related to the provision, installation, and the upkeep of their equipment and network connectivity.

Table 3

In countries such as China, where competition is intense, wholesale operators such as GDS Holdings Ltd. and Chindata Group Holdings Ltd. generally offer contracts at a fixed price with limited or no escalation. This exposes the companies to inflation risk.

Operating costs for data centers in Asia-Pacific are set to rise. Our economists expect inflation to be 2.1%-2.4% per annum over 2024-2026. In countries such as India the rate may climb as high as 4.5%-4.6%.

Utilities (including power) costs are the largest operating expense for wholesale/hyperscale data centers. Staff costs are also a big expense for retail-colocation entities.

Most retail colocation does not have direct cost pass-through. That said, some of the operators embed options into their terms, allowing them to raise prices if power costs rise above a threshold. Otherwise, the shorter contracts typical in this space allow operators to reprice more frequently to reflect higher costs, if feasible, given competitive pressures.

Why do we view obsolescence to be a key risk?

New technology and consumer preferences are constantly changing the types of workloads and the requirements of the facilities to carry out those workloads. Entities must refit data centers to match demand changes; this poses operational risks.

The shift to hyperscale data centers due to cloud migration will squeeze demand for small and midsize wholesale data centers. These smaller facilities often cannot meet the needs of cloud service providers; some enterprise customers may move to cloud-service providers, causing customer churn. The rise in high-performance computing, especially in generative AI and machine learning, may pressure operators to upgrade data centers, adding cost and execution risk.

Operators may face other limitations, including a lack of access to additional water for cooling, and power supply. Their buildings may also be ill-suited to the requirement of high-density data centers.

Many operators are now implementing flexibility in their designs, involving reserve infrastructure and land. We would consider the degree of this so-called future-proofing in our assessment of the asset quality of data center operators, among other factors such as location and specifications.

Are data centers vulnerable to cyberattacks?

Yes. Cyberattacks can threaten information security and the availability and performance of data centers. According to data provided by Guidewire, a third-party researcher, cyber risks in Asia-Pacific are just as high, if not higher, than in the U.S. or Europe. A successful cyberattack may inflict significant reputational and operational hits on operators and their tenants (see "Asia-Pacific Corporate Cyber Risks: What You Don't Know Can Hurt You," June 10, 2024).

In June 2024, a ransomware attack shut down a data center owned by the Indonesian government, resulting in disruptions of public services including digital services for immigration, and the loss of information.

How do power availability and green targets affect data center operators?

Power availability and reliability are the top constraints for new supply in the region. For instance, Singapore has limited the building of new data centers, due to the scarcity of land and power. China has also limited new builds in regions with power and water shortages.

Increasingly, we expect to see more data centers powered by renewable energy, even if that may be less prioritized during this initial rush to build facilities.

Chart 2

Data centers with high green standards and access to renewable energy have the upper hand in contract negotiations. Most major economies in Asia-Pacific and large enterprises have established carbon reduction targets. As such, governments are increasingly turning to more sustainable service providers. Enterprises such as the major cloud-service providers (i.e. Amazon Web Services Inc. and Alphabet Inc.) have also set ambitious carbon reduction targets.

The more efficient data centers are typically better placed to manage regulatory shifts. In Chinese cities such as Shanghai, regulators are prompting the closure of some older facilities where power usage effectiveness exceeds 1.7. In May 2024, Singapore announced guidelines requiring greater energy efficiency at existing data centers, and to increase use of green energy.

Table 4

Appendix

Table 5

Editor: Jasper Moiseiwitsch

Digital Designer: Evy Cheung

Related Research

• Approaches For Rating Data Center Financings, Aug. 1, 2024
• The Four Main Approaches For Rating Data Center Financings, June 13, 2024
• Asia-Pacific Corporate Cyber Risks: What You Don't Know Can Hurt You, June 10, 2024
• Data Centers In South And Southeast Asia: Balancing Risk And Reward, Sept. 10, 2023