Any data center facility worthy of its name should develop a complete roadmap and sustainability strategy. As sustainability increasingly becomes a core topic of discussion in the data center industry, several elements simply cannot be overlooked in the pursuit of a “greener grid.”

At the heart of any sustainability strategy is a clear understanding by the board and executive team of “how to take action to achieve science-based net-zero emissions targets by 2050.” The strategy should contain quantifiable, concrete targets rather than vague, empty statements. Typically, this should cover how the organization will reduce greenhouse gas (GHG) emissions, lower water consumption, and improve energy efficiency and the share of renewable energy.

As things stand, many data center operators have set up dedicated sustainability teams within their organizations——which is undoubtedly a positive development. However, these teams must work closely and in step with other functions such as operations. Their work must never come at the expense of data center reliability, resilience, or performance——after all, these are not core areas of expertise for sustainability teams.

Of course, for an energy-intensive facility like a data center, the use of renewable energy is undoubtedly one of the key aspects to consider. This topic falls under Scope 1 and Scope 2 of greenhouse gas accounting frameworks, and its quantification is relatively straightforward. In practice, however, this aspect often comes with numerous challenges.

Power Purchase Agreements (PPAs)

There has been widespread debate in the industry over whether the heavy acquisition of “Power Purchase Agreements” (PPAs) by various organizations is justified. As an industry insider who readily admits to being a “skeptic,” I cannot help but ask: “Does this power actually exist and is it actually available?” and “Has this power really been fed into the grid that supplies the data center?” In most cases, the answer is no——that power has in fact never been delivered into the grid.

There is currently no way to distinguish “green electrons” from electrons of other origins. Electricity is simply electricity; the power we draw from the grid is in fact a “stew” of many different sources. The problem arises when the grid's generation mix is dominated by high-carbon energy sources such as coal (a highly carbon-intensive fuel)——and especially natural gas, which often accounts for an even larger share. If high-carbon sources make up as much as 75% of the grid mix, then even importing some low-carbon energy from neighboring countries will only raise the grid's low-carbon share by a few percentage points.

Solar Energy

As a renewable energy source, solar power is undoubtedly a reliable source of electricity, but its application is not without preconditions. To be truly effective, any solar power project must be tightly coupled with an efficient “battery energy storage system” (BESS). After all, solar systems cannot generate electricity at night; the power produced during the day often must be stored first for nighttime uses such as charging electric vehicles. That power can also serve “parasitic loads” (i.e., non-critical electrical demand) such as lighting. And even when that power is fed back to the grid, we still repeatedly hear the same question: does the grid have the capacity to absorb this additional electricity?

Another way to leverage renewable energy is to disconnect the data center from the public grid entirely. Some data centers have chosen to transform into “standalone facilities,” building on-site combined heat and power plants or adopting gas fuel cell systems. Notably, while the latter still requires connection to the natural gas network, the reliability of gas supply is typically far higher than that of public grid power. In EMEA and the United States, successful cases have shown that both power models can effectively utilize by-products of generation (such as waste heat), achieving highly efficient use of energy.

Wind Energy

Not all “wind” is suitable for power generation——Malaysia is a typical counterexample. In other words, although wind turbines require a minimum wind speed of 4 m/s to start generating, the national average wind speed in Malaysia is only 2 m/s; in fact, wind speeds in the vast majority of the country's coastal and offshore areas never reach commercially viable levels. According to some research reports, only a few regions such as Kuala Terengganu, Mersing, and Kudat may be exceptions.

Can We Improve Facility Operational Efficiency?

In recent years, the operational efficiency of data center facilities has undeniably improved. For facilities to improve efficiency further, the only path is to adopt “adaptive power provisioning” schemes that precisely match capacity to actual power consumption demand.

When a capital asset (such as an infrastructure component) is approaching the end of its service life, we recommend replacing it with newer, more energy-efficient equipment. When selecting new equipment, size capacity to actual current needs rather than overly optimistic projections of future demand. Also, make sure you have stocked biodiesel for your emergency generators——but before that, do you actually need a generator set? Can you accept a 99.99% availability level, or must you meet the extremely demanding 99.999% requirement?

Has the PUE Metric Earned Its “Retirement”?

As the saying goes: “You cannot manage what you cannot measure.” This raises another pressing question: what effective metrics are actually available today for measuring data center operational efficiency?

Recently, more than one article has argued that “Power Usage Effectiveness” (PUE) has outlived its usefulness as an efficiency metric. Here, we need to be clear: PUE never was, is not now, and never will be a true “efficiency metric”! PUE merely reflects the additional power the facility itself consumes (i.e., everything beyond the baseline “1” in the PUE formula) on top of the IT equipment power load. The lower the ratio, the less power the facility itself requires; however, this is a metric highly susceptible to many factors, such as load levels and ambient conditions——just two examples among many.

First, PUE must be understood in the specific context of “baseline reporting.” It should be a rolling average based on data from the past 12 months, calculated in accordance with common international standards——such as the ISO/IEC 30134 series or the EN 50600-4-X series. Once you start calculating strictly in line with these standards and prepare for audits against them, you are on the right track to establishing “baseline” data for your facility; on that basis, you can clearly identify where your facility still has room for improvement.

PUE itself is not at fault……the real problem lies with its users and the organizations——who either fail to apply the metric correctly or have never truly understood what PUE really is.

PUE is merely one of many metrics, designed to help you build a “panoramic picture” of your facility's operational performance——it is by no means a tool for making simplistic side-by-side comparisons of your facility with others! What possibilities exist?

Can we make the grid greener, or make data center facilities more environmentally friendly? The answer is yes. But it requires facility owners and operators to be creative about how they source power and what uptime they can offer. This does not mean virtual PPAs are the only way forward. There is even a view that PPAs exist purely as self-serving “greenwashing.” Green hydrogen cannot yet be produced at scale, and the associated distribution methods will not become widespread for at least another decade. Nuclear power has always been a viable option, and several countries in Southeast Asia are considering it; however, given regulatory approval processes and other concerns, nuclear solutions can probably only be regarded as a medium-term option.