power

This page considers whether a wired economy necessarily uses (or merely wastes) less energy.

It covers -

• introduction - how much power does it take to run the net?
• studies - major works on the energy consumption of data centers and personal computers, the net and the broader digital economy

     introduction

Anxieties about the energy requirements of the net and the digital economy are evident in claims that -

• the proliferation of online devices has "staggering implications for the thermoelectrical power complex"
• around 50% of US electricity production will be consumed by "the Internet and E-commerce activity"
• running and cooling corporate servers accounted for 1.2% of total US electricity consumption in 2005
• the power used by US personal computers, networking gear and the telephone network in 2006 was around 350 billion kWh a year, representing 9.4% of total US electricity consumption
• the power required to maintain an avatar (such as than on Second Life) is greater than that consumed by the average person in Brazil (eg here)

A salient statement is The Internet Begins With Coal: A Preliminary Exploration of the Impact of the Internet on Electricity Consumption, a 1999 study by Mark Mills for the Greening Earth Society (a US utilities advocacy organisation). Another is Nicholas Carr's comment here.

     energy

How much energy is used to run the internet (cables, servers, personal computers, other devices) and more broadly power "the internet economy"? What is the rate of growth?

The answers to those questions are unclear.

As we have indicated in highlighting some studies below there is major disagreement about base data and projections. Although devices are broadly becoming more efficient, there are more of them within developed economies - access to a single phone, for example, no longer suffices - and growth of emerging economies is being reflected in uptake of phones, personal computers, servers, televisions and radios.

It is sometimes claimed that PDAs and similar devices are innately 'green' because they are not drawing power from a grid or generating thermal pollution. However, an assessment of their overall impact might include costs involved in battery production.

What is the impact of the domestic personal computer and non-commercial uses such as burning CDs of fileshared music? There are few comprehensive audits. The average UK household is claimed to have 23 lightbulbs in use as of 2006, forecast to increase to over 26 bulbs by 2020, with the same household owning 2.4 televisions as of 2004. Electricity use in an average Australian household as of 2001 was claimed to be -

use water heating fridge/freezer air heating/cooling lighting audio/video cooking washing and ironing pool pump other

% 33 20 14 8 7 6 3 3 6

Adding an internet fridge, PDA or mobile phone or two is thus unlikely to have a fundamental impact. In 2006 one report in the UK claimed that "boiling excess water" (67% of people supposedly put "far more water than necessary" in their kettles) used enough electricity to power every street light for seven months.

In 2005 the ABS indicated that a small number of business entities accounted for most industrial use of energy, with the 100 largest enterprises consuming 1,081.9 PetaJoules -

cohort Top 100 Top 250 Top 1,000 Top 5,000

% 55.2 62.2 70.7 79.0

Of the largest 250 energy users, 47.2% are in manufacturing, 20.6% are in mining and 10.4% are in the transport sector. 132 of the largest 1,000 energy users are in the mining sector, with 367 are in the manufacturing sector.

The UK Energy Saving Trust complained in 2005 that office workers who leave computers on standby cost British industry £123.2 million per year in energy bills, with home computers waste £41 million per year, with the "standby culture" supposedly responsible for one million tons of carbon being pumped into the atmosphere each year.
The Independent fretted that UK home PCs "emit 220,000 tons of carbon dioxide when left on standby", an emission that of course occurs in the power station rather than direct from the desktop.

Some analysts have suggested that much of the energy used by server farms or web hotels is attributable to the cooling requirements of those facilities, rather than power to keep the hard drives spinning and signals going out of the building. Jennifer Mitchell-Jackson suggests that server farms across the US used no more than 0.12% of all US electric power at the end of 2000, in contrast to claims that the growth of server farms in and around Seattle would require around 1,100 megawatts a day (roughly the amount of power used by the entire city, including manufacturers such as Boeing).

In 2006 an Ask.com representative claimed that the five leading search companies were resonsible for some 2 million servers. George Gilder inferred that power for running and cooling those servers was around 2.4 gigawatts, hyped as "an impressive quantity of electricity". It is around half the daily requirements of Las Vegas but, we note, less than the power required to produce aluminium cans, aluminium foil, alfoil beanies and other necessities of contemporary life.

The experience of Seattle - and other hubs such as New York, where a projected farm was claimed to have double the power requirements of the former World Trade Center towers - is not typical of most of the US or other parts of the world. That is demonstrated in Matthew Zook's 1998 paper on The Web of Consumption: The Spatial Organization of the Internet Industry in the US - illustrating how supposedly 'spaceless' new economy industries cluster in specific locations - and Manuel Castells' The Informational City: Information Technology, Economic Restructuring & the Urban-Regional Process (Oxford: Blackwell 1989).

Does the net mean that we are going to run out of energy? Arguably not, with the real questions instead relating to the sourcing and cost of energy supplies, the location of power stations and responsibility for externalities (whether they're local such as the disposal of ash from a coal-fired power station or the contentious greenhouse effect).

John Laitner's 2005 Economic Policy Models and Alternative Future Scenarios: Decided Room for Improvement (ppt) notes that since 1970 energy efficiency (ie improvements in production, distribution and use) has met 75% of new energy service demands in the US.

     Studies

The landmark 1999 The Internet & Global Warming report by Joseph Romm, Arthur Rosenfeld & Susan Herrmann of the Center for Energy & Climate Solutions claimed economic growth of 8% in the US during 1997-98 and noted that energy consumption grew by only 1% rather than the expected 10%. That difference was attributed to the new economy and has been the basis of claims that 'being online' will result in substantial systemic reductions in energy demand.

Jay Hakes' lucid 2000 The Potential Impacts of Computers and the Internet on Electricity Consumption disagreed, attributing lower demand in 1997-98 to an unusually mild winter and commenting sensibly that

it is too soon to come to any conclusions as to the precise path of electricity use resulting from internet and internet-based commerce.

As noted above, the 1999 The Internet Begins with Coal: A Preliminary Exploration of the Impact of the Internet on Electricity Consumption report by Mark Mills - echoed in a Forbes polemic by Mills and Peter Huber - estimated "internet related" electricity use at around 8% of all US electricity use in 1998 and growing to half of all electricity use in the current decade.

The report was produced for the Greening Earth Society - one of the coal industry lobby groups fighting the 'carbon wars' - but there have been similar claims from figures such as George Gilder. Those claims are often uncritically cited or distorted: the San Francisco Chronicle for example trumpeted in 2006 that "George Gilder projects that Internet computing will soon require as much power as the entire U.S. economy did in 2001" and the Wall Street Journal pronounced in 2006 that

To satisfy their power needs, Internet companies are exploring options ranging from building facilities in former defense bunkers - which already have rugged grid connections - to plunking themselves down near hydroelectric plants to get a slice of the inexpensive power. Anticipating demand a decade from now, some executives even are mulling whether proximity to nuclear-power plants could be a plus.

Unsurprisingly, the Internet Begins with Coal report was criticised as fundamentally flawed.

A 1999 critique (PDF) by the Lawrence Berkeley National Laboratory (LBL) for example suggested that the figures should be reduced by a factor of eight. LBL analysts estimated the annual electricity consumption of all the office and network equipment in the United States at about 74 terawatt hours. That was 2% of national consumption, rising to 3% if the cost of manufacturing the hardware was included.

Criticism has not, however, inhibited claims that Silicon Valley was responsible for the Californian power crisis or that internet hosting facilities ('server farms' or 'web hotels') guzzle more juice than some US states. One inference has been that there is an emerging power crisis in the US and other counties, thanks to the web, so restrictions on nuclear plants, inefficient coal-fired power stations and other nasties should be reduced.

The LBL Information Technology & Resource Use (PDF) study by Jennifer Mitchell-Jackson assessed energy use by data centres, explores why most estimates are significantly too high and supplies substantive rather than anecdotal figures for five facilities. The study drew on her Masters thesis (PDF) of May 2001 regarding electricity used by data centres. The thesis supplies measured top down (billing data) and the bottom up (counting equipment and measuring or estimating actual power used for each piece of equipment, then adding it up) data. Overall, the best estimate of power used by US centres is under 0.12% of all electric power consumption at the end of 2000.

The study reflects previous LBL research about the impact of information technology on resource use. In 1995 the Lab published a comprehensive assessment of power used by commercial-sector office equipment (PDF). It offered a point by point rebuttal (PDF) of Congressional testimony by Mills, during an inquiry that featured suggestions that the net - like photocopiers - should be turned off at night. Measurements in the second major assessment (PDF) of office equipment energy use released in June 2001 were consistent with forecasts in the 1995 study.

The new report suggested that total electricity used by all office equipment in the US was around 2% of all electricity consumption. Power used for all telecommunications, network and office equipment (including electricity used to manufacture the stuff in plastic boxes) accounted for around 3% of total US electricity consumption. Commercial sector office equipment electricity use is within 15% of that predicted for 2000 in the 1995 report, with the difference being attributed to by more people leaving their computers and printers on at night than envisaged in 1995.

In 2007 Jonathan Koomey's Estimating Total Power Consumption by Servers in the US and the World (PDF) offered one estimate, suggesting that

Total power used by servers represented about 0.6% of total U.S. electricity consumption in 2005. When cooling and auxiliary infrastructure are included, that number grows to 1.2%, an amount comparable to that for color televisions.

His Network Electricity Use Associated with Wireless Personal Digital Assistants (PDF) featured an estimate that power requirements for the US phone system were 3.8 billion kWh per year. Koomey questioned David Sarokin's suggestion that electricity consumption for the internet in the US was 350 billion kWh per year, with global consumption of 868 billion kWh per year.

As the following page notes, attempts at modelling the broader impact of the net or electronic commerce have been contentious, given disagreement about basic definitions, the muddiness of much data and questions about extrapolation.






    next page (ecologies)