We often think of the road to green and the road to LEED® as being one and the same. However, if your destination is a data center, mapping the current requirements reveals that the true green path may bypass this popular certification.

My first car was a ’70 Ford Fairlane. I bought it from a little old lady for $500 back when gas was $1.14 a gallon. It was a white four-door with a massive bench seat about the same size as my grandmother’s divan. It was no chick magnet (or maybe it was and I demagnetized it) but it was special because it was my first car. But an old lady’s sedan can only satisfy a 16-year-old boy for so long. So I sold it to a buddy of mine.

My buddy proceeded to try to convert the white beast into a hot rod. He painted it competition orange, bought some fat tires, and jacked it up. But you know what? Underneath it all it was still an old woman’s car with an undersized V-8 and an automatic on the column. It was never going to be what he wanted it to be. To coin a phrase, he had merely put lipstick on a pig.

Now the reason I take you down my memory fast lane is because in the last year or so, I have been approached by more than one IT enterprise wanting me to promise them a LEED® certified data center. In fact, LEED questions have, in many cases, trumped reliability queries in some interviews. And this concerns me. Because I have to tell you, that no matter what you do to green up a data center, it is still a pig … an energy hog to be precise … wearing a lovely shade of emerald lipstick.

Now Stop Right There!

“Dickens!” you might exclaim, “You green-washed hypocrite. You flipping flip-flopper. You moan for the last few years about sustainable design and now you jump the green ship just because it’s hard?”

Before anyone jumps to the conclusion that I am either bashing LEED or discouraging the greening of data centers, please take a seat. I am an advocate for LEED, and I encourage every designer to approach data centers like any other building: with a mindset to make it as energy efficient and sustainable as the program, budget, and mission will allow. In fact, it is even more important to maximize the efficiencies of data centers because their carbon footprint is so much larger than an office or commercial building with the same physical footprint.

But the problem with buying into the conventional wisdom that a green data center must be LEED-certified is that greening the design does not necessarily ensure certification. And worse, too often the pursuit and eventual abandonment of LEED certification can become a distraction from the ultimate goal of doing the right thing by Mother Nature.

There’s a good reason why, as I write this article, there are only two LEED NC-certified data centers out there. It isn’t because the design community lacks courage or vision, but rather because the laws of physics and the cold, hard checklist calculus just keep making it clear that you can’t get there from here.

Past Is Not Always Prologue

Many may say that if it’s been done in the past, it can be done again. They are right, and someone may do it again. But since those two data centers were designed, the deck has been incrementally stacked against buildings with large process loads.

The two data centers that have been certified were both certitified under earlier versions of LEED NC. And while you may have seen some hyped up reports of a Platinum pre-certified data center from Advanced Data Centers in August, that was under the Commercial Spaces rating system (CS 2.0), which has very few points that actually relate to energy efficiency. Ken Brill of the Uptime Institute nailed it when he said, “It’s sort of irrelevant for a data center.”1

Under the earlier version of the standards, unregulated process loads such as those found in data centers were not included in the energy model. Also, you were allowed to subjectively assign building plug loads, computers, printers, etc., with no ties to the building’s total energy costs. What this meant for the modeler, and subsequently the design team, is that the loads they had no control over (process and plug) were either eliminated entirely or could be minimized based on an individual’s interpretation of diversity.

In turn, and quite logically, the designer was able to focus on those load components that he actually could address: envelope, lighting, HVAC, and water heating. The relevance of this is that now to get a nominal 35% reduction in energy costs, the designer had perhaps as much as 90% of the total regulated load to work with. More precisely, to get the 35% overall reductions, one only needed to get a 39% reduction in the 90% they could actually manipulate.

100% - (90% x 0.39) = 65%

But all of that changed under NC 2.2 and 90.1-2004. First, the USGBC recognized that folks could game the plug load. The smaller the plug load, the larger the percentage of the total you had to work with, and the easier the targets were to hit. So they set the plug load floor at 25% percent of the total energy costs for the building. Now, for any building, data center or otherwise, the designer was left with only 75% of the total. So applying our previous 35% goal, the increases in efficiencies in the loads you could control jumped from 39%  to 47%.

100% - (75% x 0.47) = 65%

And if that weren’t enough, now 90.1-2004 Appendix G (I think the “G” stands for “God help us!”), said that the process loads must be equal and in both models. Ouch. Let’s say you have a data center where the process load is actually 50% of the total (a relatively low percentage for a true data center). Now you have to cut envelope, HVAC, lighting, and DWH load by 70%!

100% - (50% x 0.70) = 65%

You just turned the ratio upside down, my friend.

The Bar Goes Even Higher

Then in 2007, the USGBC, in its quest to challenge the industry (a good thing overall), bumped the Energy and Atmosphere Minimum Energy Performance prerequisite from merely meeting ASHRAE 90.1 to beating it by 14%. (Quick note: Under V3.0 they are proposing dropping down to 10%).

This hit me in the early stages of a design where the process load for my building was already approaching 70% of the total. Now instead of meeting 90.1, and likely beating it by a little, I was mandated to beat it by 14%.

So let’s run that familiar equation one more time:

100% - (30% x 0.47) = 86%

I’m good, but I couldn’t for the life of me figure out how to cut the energy use associated with regulated loads by 47% on a building with an already efficient envelope, hardly any glass to start with, a low occupancy, and very few lights I could play with. On top of that, ideas like natural ventilation and airside economizers on a data center floor were just non-starters for this client.

So you know what happened? Well, of course you don’t, you weren’t there. But I will tell you: the client bailed on the green targets altogether. Oh, I’m not saying that energy efficiency was no longer a mandate, but as an overriding theme, once LEED was off the table, from a sustainability perspective, the design team fragmented.

Now you may say shame on us for letting that happen, but in the throes of a design when you are barreling toward a deadline on a building where reliability and budget are drivers, if halfway through the process your green paradigm gets tossed, it’s hard to refocus, especially when it was a secondary goal to begin with.

Q & A

The adjustments made to NC 2.2 are punitive and will likely discourage some from pursuing sustainable data center designs, but maybe I’m just a pessimist. Maybe I have become jaded on green (pun painfully intended). Has anyone asked the powers that be for a fair adjudication?

In February 2007, an official interpretation and request for relief from this onerous requirement of equal process loads was submitted to the USGBC. In so many words, they acknowledged the challenge but offered no sanctuary2. Again in July of 2007, the same question was asked, and the USGBC stood by their guns3. In both cases these were data center designs hamstrung by the new requirements.

In August of last year, yours truly submitted an interpretation request to ASHRAE regarding the very applicability of 90.1 to data centers in the first place, since they are primarily buildings dedicated to a commercial process (a specific exception under 2.3.c of the standard). ASHRAE responded, stating that the building envelope and systems must meet all requirements of the standard with the exception of the “space conditioning system installed specifically to support the Datacom equipment.4” I found no comfort in this response because it appeared to be taking even more stuff out of my control, further increasing the percentage I must bleed from the remainder. Just FYI, further clarification has been requested from the Committee.

And one more report to belabor the point, and possibly confuse. The EPA’s Report to Congress on data center energy efficiency5 states, “The current ASHRAE standards, however, exclude energy that is primarily used for commercial, industrial, or manufacturing processes. Thus ASHRAE 90.1 excludes data center heating and cooling (which mainly cools equipment, not people) and IT equipment (ASHRAE 2004a, Section 2.3). However, ASHRAE is developing guidance to address aspects of data center design … and plans to evaluate whether data centers could be included within standard 90.1.”

Don’t you love it? According to someone at ASHRAE who must have spoken to someone at the EPA, ASHRAE is still trying to figure out if 90.1 could or should apply to data centers. In the meantime, ASHRAE tells me that it does apply and LEED says I have to meet and beat it.

The Exact Science Of Approximation

The last thing I will say about LEED and data centers … well, I won’t even say it. I will let the New Buildings Institute Energy Performance of LEED for New Construction Buildings Report6 do the talking. A key finding on high-energy use building types like data centers was that “the alignment between predicted and actual energy use was very poor, even on average. In fact, on average these buildings used nearly two-and-a-half times as much energy as was predicted during the design phase.”

The report went on to state that “this discrepancy suggests the actual performance characteristics of these building types are not well understood by the design community … It is clear there is a need for significant additional research into the performance characteristics for these building types and for direct feedback to the design and owner community. The data also suggest LEED may need to re-evaluate how these project types are treated with respect to energy performance achievement.”

The report goes on at length regarding plug loads in general and how they were treated prior to NC 2.2 and basically concludes that modeling was all over the place. So there are no apples to compare our new NC 2.2 apples to. It looks like some of the apples are oranges, some are coconuts, some are … you see where I’m going here.

So What's My Point?

I have used almost 1,800 of my 2,000 allotted words for this column to bash LEED and demonstrate that it is really the wrong tool to apply to data centers, at least for now. I have cast aspersions at previous LEED-certified data centers and may have implied that their designers gamed the system. In reality, I think no such thing. I just am saying the rules have changed dramatically.

You might think that I think no data center can ever be certified under NC, and I would tell you you’re wrong. There will always be creative and dedicated design teams that will make LEED a priority and find ways to get there. Especially when companies like Google and Microsoft are footing the bill for more aggressive and progressive facilities.

But most of us don’t work for Internet giants. Data centers are popping up all over the place, and their sizes and Watt densities are all over the map. Most are for clients and owners whose primary business isn’t IT - folks whose facility know-how is probably commercially or institutionally base; people who operate office buildings, college campuses, primary, and secondary schools, etc. These are the very people who have fallen in behind LEED, and for good reason.

The purpose of this article is to educate and enable you to intelligently and convincingly redirect the discussion of a green data center away from the LEED model, and instead to one that is more organic and project-specific. LEED may actually end up being right for your project, but it should be an outcome, not a driver, in the discussion. The destination should always be a green data center. I’m just saying that as the ones who are responsible for guiding the energy discussion, we need to start from a different perspective.

A green data center may in fact be an oxymoron, but that shouldn’t stop you and your team from creating the most environmentally responsible oxymoron possible. So admit to your client that their building is in fact an energy hog and not automatically a candidate for LEED. But then, I say sidle up to that pig, put on that green lip gloss, and with intelligence and conviction … go for it … and kiss her like you mean it. ES


1. Turner, Melanie, “$100M data center rises,” Sacramento Business Journal, June 13, 2008.
2. USGBC, Credit Interpretation Rulings, Energy & Atmosphere: Optimize Energy Performance, date of CIR 2/28/2007, date of ruling 3/23/2007.
3. USGBC, Credit Interpretation Rulings, Energy & Atmosphere: Optimize Energy Performance, date of CIR 7/10/2007, date of ruling 8/13/2007.
4. Interpretation IC 90.1-2004-X OF ANSI/ASHRAE/IESNA Standard 90.1-2004, Energy Standard for Buildings Except Low-Rise Residential Buildings, date approved –April 12, 2008.
5. EPA, Report to Congress on Server and Data Center Energy Efficiency Public Law 109-431, August 2, 2007.
6. NBI/USGBC, Energy Performance of LEED for New Construction Buildings, March 4, 2008.