What's at Stake in this Argument?
The reason why those concerned about the development of sustainability (ie Lovins and Romm) are so actively promoting the internet as a source of energy efficiency has recently become clearer. All the debaters participated in a US Senate Inquiry into energy and IT trends, the outcome is now evident in the Bush Administration's Energy Policy.
   On May 1st, Dick Cheney, head of oil-services company Halliburton Inc. before becoming vice president, claimed that the US was facing an Energy Crisis. Drawing on the same research that Mills used, Cheney claimed that the US's rising energy consumption will require one new electricity-generating plant a week for 20 years.
  "'America's reliance on energy, and fossil fuels in particular, has lately taken on an urgency not felt since the late 1970's,' Mr. Cheney said. 'Without a clear, coherent energy strategy, all Americans could one day go through what Californians are experiencing now, or worse.'… Mr. Cheney dismissed as 1970's-era thinking the notion that 'we could simply conserve or ration our way out' of what he called an energy crisis…He said he would oppose any measure based on the premise that Americans now 'live too well' or that people should 'do more with less.' "The aim here is efficiency, not austerity,' he said. 'Conservation may be a sign of personal virtue, but it is not a sufficient basis for a sound, comprehensive energy policy.'" (New York Times "Cheney Promotes Increasing Supply as Energy Policy", May 1, 2001)
   Oil exploration in Alaskan national parks, construction of nuclear power stations, and a relaxation of power generation pollution controls are all justified to sustain the power supply to the US's New Growth Economy.
   To counter this appropriation of the alarmist position, sustainability advocates have been forced to adopt an optimistic rhetoric. Lovins and Romm are insistent that there is no energy crisis and that the move to more sustainable energy policies is only a small step, in a direction the economy is already taking. Thanks to IT, energy demand reduction is underway without much pain, so that any increase in demand that does occur can easily be met by the incremental phase in of decentralized renewable energy supply technologies. There is no need for anti-environmental emergency measures, simply the sustainment, and perhaps enhancement, of existing progressive policies.
   The wider point that these sustainable developers are making is that it is possible to 'decouple' economic growth from ecological harm. The internet is seen as a major way to 'decarbonise' the economy, ensuring the continued improvement of living standards whilst reducing the risk of climate change. Environment Ministers of developed nations with pro-Kyoto electorates, especially Australia, are nervously adopting this argument, waving around recent figures suggesting that increases in greenhouse gas emissions are no longer keeping pace with economic growth rates. Romm's principle point in his presentation to the US Senate was that in 1997-8 the US economy grew by 4%, but the nation's energy consumption and greenhouse gas emissions hardly grew at all. Energy intensity, measured in energy consumed per dollar of gross domestic product, is declining, ie improving. Sustainability is no longer an anti-market revolution away, but an already-arriving market evolution.

The Resulting Relational Picture of the PC
Apart from the rhetorical reversal that has occurred (the conservatives want a radical energy supply change and the conservationists want to conserve the trends of the energy supply status quo), the debate is noteworthy in that it demonstrates that a more relational approach to sustainability issues is emerging. Either side in this debate is only prepared to be as relational as is required to shore up their positions, but, as they urged each other on, the overall result is a more comprehensive picture of the ecological consequences of IT. As can already be seen however, neither party is prepared to fully embrace relationality: a residual desire for rational answers ('what is the state of affairs?') continues to obstruct a more constructivist disposition ('how can we make multiple interventions into the state of affairs?').
   To get a sense of the relational picture that emerges from this debate, let's go through some of the details.
   The key issue is: what is a PC? Since there are many (servers, mainframes, desktops, laptops, palmtops) analysis requires that limits be set through the notion of the 'average' model. Mills originally made the relational point that a PC is more than a processor and monitor; there are also modems, printers, zipdrives, cd-burners, etc. However, in the media version of his research, Mills foolishly claimed that the average PC and all its peripherals use 1000W. This would only be the peak demand for a large professional desktop publishing PC whilst printing. Lovins and Romm claim that the average PC uses around 150W (with more than half used by the screen), and ignore the energy consumption of peripherals: "Printers and peripherals tend to be spread over a great many users and don't increase this average very much." (Joseph Romm, "The Internet Economy and Global Warming: A Scenario of the Impact of E-commerce on Energy and the Environment") They justify the choice of this averaged model by pointing out that laptops tend to have increasingly lower wattages. Mills counters by pointing out that increased size and number of PCs and peripherals are outstripping any improved energy efficiencies. Koomey attests to this, having predicted in 1995 that the average wattage for a display would be 63W, whereas it is in fact 85W.
   Part of the definition of a PC is its 'duty cycle', that is, the number of hours the average PC is in operation, consuming energy. Mills assumed that PCs are left on 24 hours / 7 days. Koomey's research indicated otherwise: 9% of household PCs were left always on, 19% were run 16-40 hours a week, 49% 2-15 h/wk, and 23% < 2 h/wk. However, Koomey also attributed the greater than predicted energy consumption levels of IT in 1999 primarily to an increase in the number of business PCs and peripherals left on overnight. As the last issue of this newsletter pointed out, it is increasingly difficult to turn PCs and peripherals completely off. Lovins and Romm also make much of the market penetration of energy management programs (such as EnergyStar compliant monitors), which do reduce the energy consumption of those PCs left on. However, our recent experience auditing the ecological impacts of a company that, amongst other activities, carries out IT training, found that more than half the staff were either ignorant that they could turn on energy management programs, or else had deliberately turned the energy management programs off.
   One of the big issues in this debate is the fact that a PC is no longer a stand-alone system, but shares the energy demand of the internet's infrastructure. Mills' work claimed that there are: 2 million U.S. Internet routers plus a further 1 million WAN/LAN routers, each averaging 1 kW and operating continuously; 4 million U.S. small website servers, each averaging 1.5 kW and operating continuously; 25k telephone central office switches, each using 500 kW (24/7); ~30k ".com" operations run on mainframes, each continuously using an average of 250 kW. These all appear to be overestimates; for instance, in relation to routers, Lovins received the following email from Cisco Pty Ltd:
   Cisco's HIGHEST-END router, the 12000-series, draws 1.5-2 kW. These are used for Internet backbones-really serious stuff. Probably several thousand have been sold over the past year or two. They are very expensive. Cisco has about 88% of market share for this size router. The biggest-selling router has been, in the past, similar to [the 2500 series, nameplate-rated at 40W, which]...is at the high end of 'typical' [intensity]. Over a million of them have been manufactured. The router of the more-immediate future, starting within the past year, is something like the 800-series baby. It's smaller and draws all of 20W. Hundreds of thousands of them have been sold and there will be a lot more.
   The last line needs to be taken with a grain of salt given the recent Tech-Wreck, lead by what Cisco has described as a 'natural disaster' in demand for its products (since such a dramatic drop was, according to management, 'utterly unpredictable').
  Usefully, Mills also insists that the energy consumption of internet infrastructure involves not only routers, servers, switches and mainframes, but also the HVAC systems keeping all this equipment cool, and the back up power systems (providing what is called 9 9s guaranteed power supply, with outages assured to be less than 0.0000001% of the time). As this newsletter has previously discussed, server farms, buildings packed with internet infrastructure, tend to need as much power for the airconditioning as for the IT itself. However, Lovins and Romm argue that to date server farms are being over serviced: "If you actually go measure, you'll see the latest data centers are on the order of 50 to 70 watts a square foot," says Lovins. "They actually have nowhere to go but down because they have terrible HVAC [heating and ventilation] design. But, there's a cover-your-ass mentality - people know they'll get blamed if they let the lights go off, but not if they waste money." The Edison Electric Institute's Steve Rosenstock concurs: "The companies building server hotels have been working with developers and asking for 150 to 200 watts per square foot. But our members have been measuring what they're actually using after installation. It's closer to 25 to 40 watts per square foot." (Jonathon Angel, "Emerging Technology: Energy Consumption And The New Economy", Network Magazine)
   There is also the embodied energy of IT equipment, especially chips, which are very energy intensive to make: currently, it takes around 8kWh to make a square inch of silicon chip, though some manufactures are starting to achieve half this. Neither side take into account the increase in embodied energy necessary to take account of end-of-life IT scenarios, whether component and materials recovery (unlikely) or landfilling (most of the time),a substantial load given the excessive turnover in IT equipment.
   Finally, there is the energy consumed by all the activities made possible by internet'ed PCs. This is an area of much contention. The central question is whether e-commerce replaces conventional purchasing practices, or merely adds to them. If it replaces cars-to- mall shopping, there is a chance that it is less energy consuming. Romm cites for example the claim the energy costs of a walk-in bookstore are $1.10 per sq.ft, as opposed to Amazon.com's warehouse at $0.56 per sq.ft. Romm makes some ridiculous claims (admittedly at the height of e-mania) about the likelihood of the efficiencies of e-commerce leading to a reduction in the demand for commercial property: a 0.7% 'internet energy efficiency' over 7 years would lead, he insists, to the elimination of one billion square feet of commercial warehouses and on-site storage at manufacturing facilities.
   In terms of transport, Romm argues that: A 20-mile round-trip to purchase two 5-pound products at one or more malls consumes about one gallon of gasoline. Having those packages transported 1000 miles by truck consumes some 0.1 gallons (and much less than that if railroads carry the packages for a significant fraction of the journey).
   However, Romm admits that the figures are not as favourable if the goods are air freighted, which is the preference in the pseudo-instantaneous and globalised internet world, and that the figures come out against e-commerce if the physical shops are nearby. As this newsletter has shown in the past, nearly every third person in your street would have to be using the same internet grocer as you at the same time for delivery to be more energy efficient than you going to fetch it.
   The primary problem with all this however is that e-commerce is not replacing cars-to-mall shopping, but merely setting itself up alongside. Consumption levels are increasing, not remaining static as one mode displaces another. This means that the energy consumption levels of e-commerce's real estate and transport, even if reduced when compared to old economy energy consumption levels, are being added to existing levels.

Consequent Relational Actions
There are many more aspects to this picture, but for now, it should be clear that this debate, despite its exaggerations on either side, has usefully expanded our sense of what a PC is in the context of energy consumption: PC = high embodied energy components + multiplying peripherals + extended operational hours + energy intensive internet infrastructure (including HVAC, etc) + additional consumption of more highly packaged and transported goods.
   As EDF has often argued in relation to LCA, the usefulness of such a picture is not in some conclusion - X is better than Y, or in this case, the internet is increasing or decreasing energy consumption - but in the multiple sites it opens up for improvements. With action at all of the levels mentioned, exponential sustainments can be made.
   In the end, it is apparent that either side of this debate is trying to sustain the status quo: current levels of consumer demand. Whether we need to increase energy supply, or promote the internet with its energy efficiencies, in either case, the foregone conclusion is that consumption levels should continue to grow. In a typically American way, neither party is prepared to confront the issue of demand reduction, since that would require making actual changes to people's lifestyles and values.

The EcoDesign Foundation is planning a regular major examination of public environmental reports and triple bottom line reports. Anyone interested in participating as a voluntary researcher should contact Cameron Tonkinwise.