Oregon data center gloom -- or not by Doug Mohney
Last week, Oregon was doing the big freakout because data centers were coming to the state and neighboring Washington state for cheap power. If data centers keep coming and fail to make their massive facilities more efficient, they could consumer nearly 10 percent of all the Northwest's energy by 2030. Lotta fuzz in "if," "fail," and "could."
"Data centers proliferate in Oregon, and power planners raise a red flag," ran in The Oregonian on September 8, based in part by a June report issued by the Northwest Power and Conversation Council. Part of the underlying beef here is that the "secretive" data center industry and utilities aren't saying much about long term plans.
Data centers are coming into Oregon because industrial power costs as little as 3.34 cents per kilowatt, with no sales tax Oregon and a rural data center tax exemption in Washington. Oregon also provides enterprise zones for data center projects that are exempt from property taxes.
The worst case scenario painted by the Northwest Power and Conservation Council would see energy consumption approaching that of region's aluminum smelters back in the 1980s, around 2,400 MW; the smelters were at 3,200 MW back in the day.
Will the worst-case come to pass? I doubt it.
On the demand side, operational expenses are always first in the mind of any company these days, with power being the biggest ongoing line item for a data center. Facebook and others are pushing to maximize energy efficiency throughout whole data center from top to bottom. There's also an upper bound to how many servers will be built and deployed within a particular region, due to geography and bandwidth considerations.
Thinking out of the box, vast and cheap deposits of natural gas throughout the U.S. in combination with lower-priced Bloom Energy fuel cells will likely start opening up competition in areas not traditionally considered for a data center. If cheap natural gas is bringing back energy-intensive manufacturing back to the U.S., certainly natural-gas fuel cells provide an opening for rethinking data center placement over the next five years.
Local regulators and politicians are already looking at building in linkages (i.e., taxes and fees) to data center energy consumption. In a down economy, local governments like to provide incentives for job creation. Data centers create high-income jobs, but not a lot of them. All those big data centers will end up being big fixtures in the community -- but also more difficult to relocate in many respects.
Finally, I always get a little annoyed at anyone trying to predict IT trends beyond 24 to 36 months. Moore's law - one of IT's gospel pieces -- has predicted a doubling of transistors in circuits every couple of years starting in 1965. Since the mid 1980s, skeptics have predicted the end of Moore's Law, crowing about running up against some technology barrier or another to derail densities. Each time there appears to be a potential for slowing, the silicon industry has figured out ways to keep on going for another seven to ten years.