Environmental watchdog group Greenpeace is taking on the technology sector again. On Wednesday, the organization released a 52-page report “How Clean is Your Cloud?,” analyzing the energy policies and data center operations of some of the world’s leading technology companies, including the likes of Oracle, HP, IBM, Microsoft, Amazon, HP, Facebook, Google, and Apple. Greenpeace’s take? With some exceptions, technology companies are mostly failing to adopt clean energy practices and are still largely relying on so-called “dirty” energy sources like coal-fired power plants to fuel their power-hungry data centers. Specifically, Greenpeace has singled out Apple, branding the company as the most reliant on coal-generated power of any firm in its survey. (In an unusual move, Apple quickly refuted Greenpeace’s analysis, and Greenpeace fired right back.)
Is cloud computing really bad for the environment? What can tech firms do to minimize their impact? And what’s Greenpeace’s agenda here?
Greenpeace’s report ranks 14 technology companies on seven criteria. The first three criteria focus on energy sourcing: how much of the companies’ power Greenpeace estimates comes from coal and nuclear sources, respectively, along with an overall “Clean Power Index,” based on estimated power demands for company facilities like data centers. The “Clean Power Index” is essentially Greenpeace’s overall ranking of a company’s energy practices.
The other four rankings are letter grades (A through F) for how well Greenpeace feels the company is behaving on four key criteria:
- Energy transparency
- Infrastructure siting
- Energy efficiency and greenhouse gas (GHG) mitigation
- Renewables and advocacy
Each of these are basically Greenpeace’s assessments of how well companies are walking the walk on clean energy. Of these points, the first (energy transparency) is perhaps the most controversial, and in fact lies at the heart of Greenpeace’s current tit-for-tat with Apple. Greenpeace wants companies to reveal detailed information about their energy consumption and sourcing, both on an overall and a site-by-site basis. That would enable advocacy groups like Greenpeace (as well as regulators and consumers) to accurately evaluate the companies’ energy policies. However, most tech firms consider this information proprietary. After all, if you build a data center and than reveal how much electricity it’s using, that gives your competitors a pretty good idea of what that data center can do. That, in turn, gives them insight into how you’ll have to invest your money for future growth and long-term operations. Most companies don’t like to lift the curtain that much on their way of doing things.
Unfortunately, that means Greenpeace often has very little data to work with for its analyses, and often has to use estimates, rankings, and third-party sources to take its best guess at how companies are doing.
So how does Greenpeace think companies are doing? Would you believe Yahoo got the highest Clean Energy Index, with a rank of 56.4 percent? And that score is essentially neck-and-neck with Dell, which came in with a 56.3 percent ranking. From there, it’s a long fall to third place, with Google earning a 39.4 percent ranking, Facebook coming in with a 36.4 percent ranking, and Rackspace (a large hosting provider) coming in with a 23.6 percent rank. At the bottom end: Salesforce, with a Clean Energy Index of just 4 percent. Oracle was also in the single digitas with a 7.1 percent ranking. From there is IBM, Amazon, and Microsoft, with rankings of 12.1, 13.5, and 13.9 percent, respectively. Apple earned a 15.3 percent.
Why did Yahoo and Dell rank well?
Greenpeace ranked Yahoo highly primarily because of the cooling designs employed in some of its data centers, and large data centers in the states of Washington and New York that can be driven by hydro power. Yahoo’s so-called “chicken coop” data center design cools servers using outside air, rather than by a more-or-less industry-standard practice of employing industrial-scale air conditioning. Even on the hottest days, outside air is usually cooler than the heated air inside servers. Typical data centers might put as much as 40 percent of their power into cooling, a cost driven up over the years by more gear running hotter in less space. Meanwhile, Yahoo says its chicken coop data center in Lockport, New York puts only about 1 percent of its power towards cooling. Yahoo legitimately deserves some applause for that.
Conversely, Greenpeace doesn’t say much about why Dell’s data center operations ranked so highly. The company only gets mediocre marks for transparency, siting, and efficiency. But Greenpeace touts the company’s preference for renewable energy and purchase of renewable energy certificates (RECs) — basically certification that Dell paid for the production of renewable energy, even if that energy didn’t go into Dell’s data centers. Greenpeace also ranks Dell as consuming less coal- and nuclear-generated power than any other company — although it’s not clear how much of that is due to REC purchases. It’s worth noting that Greenpeace’s analysis is only about data centers: Dell’s manufacturing operations and outsourcing weren’t considered.
Greenpeace also had praise for Facebook and Google as companies willing to publicly commit to renewable energy practices and supplies. Greenpeace seems to adore Facebook’s new data center in Lulea, Sweden (which is almost exclusively hydro-powered) along with its Open Compute project, which includes server designs and specifications designed with an eye towards energy efficiency. Greenpeace also lauded Google’s public commitment to energy reduction — practices that have lead Google data centers to consume as little as half the energy of a standard data center. Google also ranked highly for transparency, including publishing information on its energy procurement policies and the footprints associated with selected services.
What about Apple?
As usual, Greenpeace didn’t have very kind things to say about Apple’s data center policies, giving the company “D” rankings in everything but the infrastructure-siting category, where it earned an F. That ranking is based on Apple’s data center efforts in Maiden, North Carolina, and Prineville, Oregon: According to Greenpeace, Apple picked these locations without regard to their access to renewable energy sources, and as a result picked areas primarily served by “dirty” utilities that are 50 to 60 percent reliant on coal-based power.
Apple has responded that Greenpeace’s estimate that the North Carolina facility consumes about 100 megawatts of power is way off base: According to Apple, at full capacity the Maiden data center will draw about 20 megawatts, which Apple is mitigating by drawing from a substantial solar farm and a fuel-cell-powered facility that will eventually supply as much as 60 percent of the facility’s power. Greenpeace says, in the absence of information from Apple, it made estimates based on the size of the data center and typical industry practices — and was pretty conservative about it.
Picking on Apple’s Prineville locale also puts Greenpeace in an odd spot: the site is just down the road from Facebook’s massive Prineville data center. Last year, Greenpeace slammed Facebook for its Prineville data center, claiming that it gets nearly 63 percent of its power from coal via Pacific Power. The same is presumably true for Apple, but this year Facebook gets praise. True, Facebook gets a bit of the power for its Prineville data center from a 100 kW solar array, but a major factor in Facebook’s (and, presumably, Apple’s) decision to build in Prineville was the climate: Although summer days are hot, winters are cold and the desert climate means summer nights are cool. That means data centers can use outside air much of the year to cool servers. Low humidity means water-based “swamp coolers” are sufficient to chill air (and much of the water can be recovered) rather than industrial-scale refrigeration equipment.
[Disclosure: A long-time acquaintance helps operate Facebook’s Prineville facility, though we have only discussed it in passing.]
PUE and CUE
Setting aside the accuracy of Greenpeace’s energy estimates, the organization does make a salient point about how the data center industry measures power consumption. The industry standard metric is Power Usage Effectiveness, or PUE. PUE is the ratio of total power consumed to the amount of power needed just to run computing equipement. An ideal data center would have a PUE of one, meaning all the power consumed by the facility is used to power its gear. In reality, a typical data center has a PUE closer to two: half the power goes to servers, switches, and gear, and the other half goes to cooling, lighting, and the popcorn machine in the employee lounge. More-efficient data centers can have PUEs from 1.6 down to even 1.1.
The PUE score is useful to companies and data center operators, but Greenpeace notes that it is essentially useless for evaluating how green a data center might be. Let’s say a large, pretty-efficient data center consumes 50 megawatts of power and has a PUE of 1.5 — but it gets half its power from coal-based sources. Just down the road, there’s a 100 megawatt data center that’s almost entirely run by hydro power — but it has a PUE of 2.0. Which is greener?
It’s easy to point to the hydro center, but PUE metrics wouldn’t illustrate it. The data center relying partly on coal power is using just half the energy of the hydro data center. Using PUE scores, the coal-reliant data center is the hands-down winner.
Greenpeace argues that Carbon Usage Effectiveness, or CUE, is a better way to evaluate the “greenness” of data centers. Developed by The Green Grid, the CUE score reflects the amount of CO₂ emissions from the data center energy consumption to the amount of energy needed to power its computing gear. Essentially, it’s the carbon footprint of the computing power in a data center. And guess what? Only one company in Greenpeace’s survey provides CUE scores for its operations, and that’s the massively distributed content distribution network Akamai. Using CUE score, the hydro data center in the example above would be the hands-down winner: After all, CO₂ from hydro power are far, far lower than coal-produced power.
However, even using a CUE score to measure “greenness” can be problematic. The hydro data center in our example above is still consuming a ton of power that it’s not using for computing; that means less power for other customers of that utility, increasing power demand and potentially making the utility purchase power from less-clean sources. Similarly, Greenpeace marks down data center operators for relying on nuclear power. Nuclear power has always been controversial, but it’s worth nothing that in terms of a CUE score, it does pretty well: Estimates vary, but CO₂ emissions from nuclear power generations are broadly comparable to onshore wind farms, and generally lower than solar power generation and even offshore wind. A CUE score doesn’t take into account the 50,000-year issue of nuclear waste.
So who wins?
As with several of its previous environmental assessments of tech companies, Greenpeace is picking on Apple because it is the logical media target. For years, any news surrounding Apple gets much more attention than most other technology companies. Greenpeace gets much more mileage out of its analysis if it attacks Apple than it would if it attached Salesforce — the company that actually ranked last in Greenpeace’s Clean Energy Index.
To the extent that Greenpeace promotes discussion of issues surrounding clean energy and what drives the cloud-based services so many consumers enjoy, Greenpeace’s reports benefit the greater good. However, the more easily Greenpeace’s reports can be picked apart for inconsistencies, reliance on estimates, and making judgements on the basis of few (or no) facts, the less value they have. Greenpeace runs the risk of being the environmental organization that cried (or coughed) “wolf.”