Google, Apple, HP, all the data center folks flocking to Iceland, and other socially responsible players big and small are making various levels of commitment to relying on renewable energy for their data centers, usually accompanied by fanfare and chest thumping. Is it finally time for the rest of us to jump on the bandwagon? And what exactly does such a jump actually entail? The answer to the first question, in typical engineer-speak, is “It depends.” The answer to the second question may be a little simpler than you might have thought.

First – a quick refresher: what is renewable energy? A rather straightforward definition is that renewable energy is energy harvested from naturally replenished resources such as sunlight, moving water (waves and tides, in addition to more traditional hydroelectric power), the wind and geothermal heat. According to The Renewable Energy Policy Network for the 21^st Century, from 2008 to 2014 we have seen a global 21% CAGR in renewable energy capacity (17% CAGR for hydroelectric capacity, 44% for wind and 158% for solar); while those are impressive trend lines, total global renewable energy capacity sits around 20% of total overall energy capacity.

Methods to Secure Renewable Energy

There are three general ways to secure renewable energy:

1. Buy renewable energy from the local utility
2. Build partial on-site renewable energy capacity
3. Build complete on-site renewable energy capacity

The effectiveness of these paths will vary according to the energy source, geographic location, load demand and other variables.

Buying Renewable Energy

Buying renewable energy can be done by purchasing Renewable Energy Credits or by entering into a long-term power purchase agreement with a local utility. The former may be available to anyone willing to pay the associated premium while the latter is typically more favorably negotiated with a buyer of sufficient critical mass to fund development of some economies of scale within the utility. In either case, geography is an important factor. For example, geothermal power is part of the enticement to draw data centers to Iceland. Low-cost hydroelectric power remains important for data centers in the U.S. Pacific Northwest. The West Texas wind farms benefited from a 12 year HP power purchase agreement. Solar power might be more attractive in Phoenix than in Seattle or Chicago. Hydroelectric and geothermal power, where available, are generally cost-effective. Data centers with the option to physically locate near these resources can expect operational cost benefits over the generally ubiquitous coal, natural gas or atomic generated power. In fact, as I will discuss below, these are the only renewable energy sourcing strategies that pass normal business payback gate-keeping metrics.

Partial On-Site Build

On-site renewable energy production can be problematic with hydroelectric power or geothermal power, so we typically see this done with wind turbines or solar photovoltaic arrays. If this is a partial capacity project, then remaining capacity can either be secured from normal local utility source or from on-site gen-set capacity. Both approaches will help an organization meet carbon footprint reduction or growth mitigation goals. The off-grid gen-set strategy may support energy independence objectives but may conflict with local noise or hydrocarbon emission limits. Hydrogen fuel cells are now being entertained as a back-up especially to highly variable renewables such as the wind and the sun. While fuel cells are emission free, they are not currently renewable or sustainable unless some form of biomass energy source can be used for powering the conversion process. That is probably a good story that just doesn’t happen to be ready to be told yet.

Complete On-Site Build

Complete on-site production of the wind or solar energy and the resulting independence and carbon-neutral status may not be a realistic strategic objective for most of us yet. For example, Other World Computing in Illinois may be the first data center totally self-sufficient on its own wind-turbine energy resource. Other World Computing is a small ISP, running less than 50kW IT load. They spent $1.25 million for a wind turbine and estimate a 10-14 year payback, which may be a short-lived payback with a 15-20 year turbine life before major maintenance investment is required. They did start billing back the local utility for over-production after around six years at an average 10% rate, to contribute toward that estimated payback horizon.

The economics of solar renewable energy on-site production may be even more upside down than wind energy on-site production. Space requirements are the biggest factor today and a couple examples will illustrate this point. i/o Data Centers has invested in an 11-acre roof solar array generating 4.5MW of power. If we assume that is a hyper-efficient space with maybe 300kW expended for free cooling fan energy, that leaves us around 4.2MW capacity for IT load; however, actual delivered energy is a small portion of capacity due to night time, hours of cloud cover and various internal losses. I have seen that ratio range from 8:1 down to 5:1 from different cases studies, so we’ll assume 5:1 here just for approximate calculation purposes. At that ratio, the 4.5MW solar array could support an 800kW IT load. At 200 watts per square foot, that constitutes about 4000 square feet of data center white space, supported by nearly 480,000 square feet of solar panel-equipped roof – plenty of space left over for offices, airport parking and a ball field. Sonoma Mountain Data Center is building a 1MW 83,000 square foot solar array. Assuming a hyper-efficient 1.05 PUE and that 5:1 ratio of capacity to averaged delivered service, that 83,000 square foot array would support a 190kW IT load in a 950 square foot data center loaded to 200 watts per square foot. Finally, Apple’s 20MW solar array in North Carolina takes up about 10 acres and delivers about 42 million kW hours per year, or about a 4.8MW average peak load. At 100 watts per square foot, that comes out to about 48,000 square feet of white space served by 435,600 square feet of solar energy collection. I have not seen any ROI forecasts for any of these larger projects, but the folks at Emerson are saying their solar powered data center project in St. Louis will have about a 20-year payback.

Objectives of Renewable Energy

Except for the folks building data centers in the far Pacific Northwest, I don’t hear much ROI or payback conversation from folks tying their data centers to renewable energy. Rather, I hear strategic objectives such as:

1. Stabilize resources – long-term power purchase agreements or on-site production minimize susceptibility to external events.
2. Meet greenhouse emissions or carbon neutrality goals – Hopefully these examples of good citizenship will generate some copy-cat behavior in the industry. Some of these companies expect some residual marketing positive image benefits, but their front-end investments should lower the obstacles to entry for folks entering later and hopefully make it easier for more of us to do the right thing.
3. Achieve better energy price stability – while most costs remain higher than carbon-emission dependent energy (except, of course, some hydro and geothermal), more stable cost structures are more manageable than the regular wide swings in the traditional energy sector.
4. Achieve greater independence – umbrella for more stable resources and prices.

Should You Invest?

On the surface, it would seem that unless you can locate your data center geographically near access to inexpensive hydroelectric or geothermal power, the business case does not seem terribly attractive for renewable energy for powering the data center. However, as citizens of the planet, it may be critically necessary. All too often, the “right thing” does not get done while we’re all waiting for someone else to take the first step and ease the way for the rest of us. In this case, some pioneers have already taken those first steps and for that, we owe them a vote of gratitude for lowering the economic and technical barriers to entry.

Finally, given the cost premiums and longer paybacks associated with many renewable energy initiatives for data centers, and given that these initiatives are still in the best interest, ultimately, of all of us, a good starting point down the path of implementing a renewable energy plan is to be sure that everything else has been done to minimize the energy use in the data center. This means that steely-eyed discipline has been applied to culling all out-of-service or superfluous electronics, virtualization has maximized IT equipment utilization, all best practices of airflow management have been applied and fine-tuned, and reliance on mechanical cooling has been eliminated or at least dramatically scaled back. With all these elements under control, the step into renewable energy does not need to be so treacherous.