18 Comments

Summary:

Amazon’s web infrastructure rockstar James Hamilton writes on his blog that he’s not convinced that solar power is a good fit for powering data centers. Using examples from Facebook and Apple he lays out a compelling arguement, but he’s missing a few points.

Photo by Katie Fehrenbacher/Gigaom

James Hamilton of Amazon

Amazon’s web infrastructure rockstar James Hamilton writes on his blog that he’s not convinced that solar power is a good fit for powering data centers. He points to the two most high profile examples of Internet companies embracing solar for data centers — Facebook’s tiny solar project in Prineville, Ore. and Apple’s massive planned solar installation in North Carolina — and says that based on these projects, solar doesn’t seem to be a big win for data centers.

Hamilton points out that the 100 kW solar array at Facebook’s 25 MW Oregon data center is so tiny — contributing 0.055 percent of the facility power, he calculates — that it’s a pure marketing move.

It might run the lights in the datacenter but it has almost no measurable possible impact on the overall energy consumed. . . . As a point of comparison, this entire solar farm produces approximately as much output as one high density rack of servers consumes.

On the other hand, Apple’s 20 MW planned solar farm for its estimated 100 MW data center, is so large that Apple is clearing trees off of 171 acres to make room for part of it. Solar panel farms indeed take up a lot of space, and because the sun doesn’t shine at night, and can also be variable during the day, they only have a certain amount of usable output (Hamilton says a 20 MW solar farm has a 15.8 percent output and would yield around 3.2 MW). Hamilton crunches the numbers and says that Apple’s solar farm would have to be 24.4 times larger, at 488 MW, and on 4,172 acres, to power the entire Apple data center.

Hamilton says he’s looking to energy efficiency first, before intermittent clean power such as solar. No word on what he thinks of powering data centers with fuel cells, like Apple plans to do.

A few Internet companies have been slowly experimenting with sourcing clean power for data centers, either building their own solar farms (like Apple is doing), or investing in clean power projects (like Google does). What Hamilton doesn’t factor into his analysis, is a couple things:

  • 1). Building or buying clean power can pay off over the life of the system. Since tapping into the sun is free, it can deliver a return over many years, particularly if local energy prices rise. So the choice to go solar can also be an economic one.
  • 2). The externality of climate change. If more and more data centers are built and run off of entirely coal power, the development of the Internet will be contributing to climate change.

Hamilton, has a point: solar technology is relatively new in the grand scheme of energy, is more expensive than fossil fuel power and is variable, while data centers need baseload power. But I think solar can be part of the equation, and particularly in the way Google is doing it. Google is investing in solar power projects, like BrightSource Energy’s solar thermal farm, that are selling clean power to the local utility. Instead of building its own solar plant, Google could then build a future data center in the utility’s footprint.

You’re subscribed! If you like, you can update your settings

  1. Has anyone looked into wind power? Which goes day and night in some passes. Wrap a wind farm around a data center. You can also use the wind for cooling.

    In my mind the real improvements will be in keeping the voltage conversions down. A server really only needs 12V to run the drives. And everything else runs at 5V or less. Running the solar panels up to some industrial voltage like 480VAC and then chopping it back down to 12/5VDC is a real waste.

    Share
    1. Vera Comment Monday, March 19, 2012

      well, I’m pretty sure the sun is going to come up tomorrow, but is it going to be windy..? talk about variable yield.

      Share
      1. Ask the folks in Fairfield or the Altamont pass which already has a huge wind farm (built largely as a tax dodge, but hey it helps) Put a data center in the midst of those things and it will run around the clock.

        Share
  2. Really? Is there such a thing as a cloud groupie?

    Share
    1. web infrastructure fan boyz out there!

      Share
  3. Katie, what about Bloom Energy’s fuel cells?

    Share
    1. What about them? Haven’t gotten a confirmation on the deal from either company, but I still think this is true: http://gigaom.com/cleantech/looks-like-bloom-energy-is-behind-apples-massive-fuel-cell-farm/

      Share
  4. Vera Comment Monday, March 19, 2012

    what he’s saying is probably true, but datacenters are not a short term investment. PV efficiency will get better, as will server efficiency. We’ll never see them negate each other.. but there really isn’t a downside (he’s not paying for it).

    instead of looking at how much of the datacenters draw is offset, you should look at it in terms of carbon emissions saved.. and how much money would the yield of the array cost if you pulled it from the grid instead (don’t forget the utilities charge a higer rate during the time PVs are most productive. I have PV’s on my roof. Even in 100% cloudcover the yield only drops 10-15%)

    how many houses will 3.2MW power?

    Share
  5. Wind power is unreliable and unpredictable. There are only a few places in the US where the wind blows with any consistency (ask the Wright brothers). Additionally, there is no way to store energy produced by wind farms. Thus, all wind power is use it or lose it. Solar is not a new technology and it is still very inefficient. Ms Fehrenbacher’s two points at the end of her piece as inaccurate. Most solar arrays have only 30-year lifespans (due to the physics of semiconductors) and do not come close to penciling out. And do not forget to consider the amount of energy required to clear and prepare the solar site(as in Apple’s case) and to manufacture and transport all components to the site. As for climate change (meaning in her case, global warming caused by CO2), there has been no warming measured in over a decade (regardless of what the “models” predict). Strange, since the amount of CO2 continues to rise.

    Share
    1. “There is no way to store energy produced by wind farms.”

      Have you not heard of batteries, flywheels, compressed air, etc.?

      Share
  6. Perhaps the solar panels make economic sense if they let you shave the peak energy consumption for cooling during the hottest part of the day.

    Share
  7. Actually it is new in comparison to other energy technologies, that’s why the price is still dropping.

    Share
  8. Is current Solar technology THE answer? No. But is it a step in the right direction? Probably. If it generates the discussion, research, and development, maybe we get to something feasible. For those saying it’s not possible, just remember, nearly all technological advances were, at some point, considered impossible or impractical.

    Share
  9. These are pilot plants. Solar power is getting cheaper by leaps and bounds, and industrial-scale batteries are looking like they will become practical in the next 10 years. The next generation of solar plants will make much more economic sense, and these companies will have lots of experience they can use or sell.

    Share

Comments have been disabled for this post