In late September, while on the stump in Michigan, George W. Bush outlined his energy plan for America. More domestic oil drilling was needed, he told the crowd, because the country needs more natural gas. We also need more renewable energy and more electric power because, he said, “today the equipment needed to power the Internet consumes 8 percent of all the electricity produced in the United States.”
Bush isn’t the only one to use the 8 percent figure. Over the last 18 months, that estimate, first published in a May 31, 1999, article in Forbes by Peter Huber and Mark Mills, has appeared in reports issued by investment banks, in energy projections issued by natural gas companies and in numerous magazines and newspapers.
But is it, in fact, an accurate reflection of reality present and future?
Questions about the Internet and electricity usage gained velocity this fall after tech guru George Gilder and the Energy Information Administration (EIA) both weighed in on the topic. In September, Bambi Francisco of CBS.MarketWatch.com reported that Gilder was predicting the Internet would “eventually use as much electricity as the entire U.S. economy does today.” In late November, the EIA significantly increased its projections for future electricity usage and it named computers as a cause.
Meanwhile, California continues to be rocked by serious shortages of electricity. Last week [On Dec. 7], the state declared a Stage 3 emergency, meaning that residential and commercial users could face rolling blackouts. [insert] On Dec. 13, the power shortage became so severe that US Energy Secretary Bill Richardson threatened to hit out-of-state power producers with federal price controls if they didn’t begin shipping more electricity to California immediately. Richardson also ordered federal hydroelectric networks to boost power generation. His moves helped the state avert widespread blackouts. But the long-term prospects for California’s electric power system are unclear.
All these predictions and events have helped fuel a surge in the prices of once-stolid electric utility stocks. They have also contributed to the widespread belief that the Internet is causing big increases in domestic electricity usage. And while there is evidence to support that belief, it’s not yet certain that the Internet is causing or will cause Americans to use vast amounts of additional electric power.
There’s another problem: Huber and Mills’ 8 percent estimate appears to be wrong. All office, telecommunications, and network equipment in the country actually uses about 3 percent of the power consumed in the United States, say a group of staff scientists at Lawrence Berkeley National Laboratory. Their analysis ( http://enduse.lbl.gov/projects/infotech.html ) is supported by Steven Taub, an associate director at Cambridge Energy Research Associates ( www.cera.com ), an energy consulting firm in Cambridge, Mass. The Lawrence Berkeley scientists are “much closer to right,” than Huber and Mills, says Taub.
Questions about the Internet and electricity consumption are part of a broader debate over global warming. If Mills and Huber are right, the expansion of the Internet is contributing to the warming of the planet and is therefore damaging the environment. If their critics are right and the Net is lowering the quantity of power needed to keep the economy humming along, it should be having a positive impact on the environment.
Despite the attack on their estimate, Huber and Mills have refused to back off their 8 percent figure, triggering a rancorous debate over their motives, methods and credentials. Not surprisingly, the debate is suffused with politics: Huber and Mills are free-market conservatives who argue that the solution to looming electricity shortages is to build more big power plants. Their critics, generally speaking, are left-leaning energy analysts who favor distributed generation plants, higher efficiency products and renewable energy programs while opposing big new power plants. Increased use of the Internet, these critics argue, is increasing productivity and significantly reducing the nation’s “energy intensity” the amount of power needed to produce goods and services.
Quality vs. Quantity
Even if Huber and Mills and Gilder are wrong about the quantity of power used by the Net, the debate they have spawned is long overdue, and the issues they are raising are crucial for the Internet and the people who use it. The nation’s aging power grid is sagging under the strain of surging economic growth and the need for more reliable power. According to the Electric Power Research Institute (EPRI), America’s generating capacity has grown 30 percent over the past decade, but its transmission capacity grew just 15 percent. Over the next 10 years, EPRI predicts generation will grow another 20 percent, but the infrastructure needed to deliver that power will grow by just 4 percent.
Nowhere are the problems of generation and transmission more evident than in California. “It’s a dire situation,” says Michelle Montague-Bruno, a spokesperson for the Silicon Valley Manufacturing Group, which represents 190 companies in the region. To save power, members of the group are turning off nonessential equipment, including some lights and computers. Electricity became a key issue for the group in June when Pacific Gas & Electric was forced to cut power to about 100,000 customers after managers of California’s power system warned that the Bay area’s power grid was near collapse. Those problems continued into December, when the problem was exacerbated by cold weather, power plants idled for maintenance and a shortage of electricity that can be imported.
“There’s a perception that the high-tech industry is responsible for the boom in power consumption. That’s not necessarily accurate,” said Montague-Bruno. “It’s due to the boom in construction and the overall growth in the economy and population in California.”
Since 1998, California has licensed eight new power plants. But that new power won’t begin coming on line until next summer, which means the state’s electricity woes are likely to continue for many months to come. Some companies have begun to look more closely at on-site generating equipment including gas-fired turbines and fuel cells. Those technologies could be implemented at far lower cost than centralized power plants and would obviate the need for big investments in new high-power transmission lines. In addition, on-site generation is more efficient. When electricity is transported long distances, significant amounts of power are lost due to resistance from the wires themselves.
Deregulation was supposed to lead to cheaper, more reliable power. But so far, deregulation, now under way in California and about two dozen other states, has only confused the nation’s power picture. And all types of businesses — from manufacturers to dot-coms — are being forced to deal with questions about the availability, reliability and quality of electricity. America’s power grid was “built on a 1950’s and 1960’s design that doesn’t address the type of reliability that we need,” said Karl Stahlkopf, vice president of power delivery at EPRI.
Stahlkopf, too, believes Huber and Mills overstated the amount of power used by the Net. But he argues that much of the debate is over the wrong issue. Yes, the quantity of power used by the Net is important. But Stahlberg says Huber and Mills have been prescient in their discussion of the reliability issue, which may be a longer-term problem than questions about availability. “The nature of power usage in America is changing because all silicon-based equipment needs absolutely reliable power,” he said. EPRI estimates that power interruptions — some lasting just 1/60 of a second — are costing American businesses some $50 billion each year. Sun Microsystems has estimated each minute of power outage costs it $1 million, said Stahlkopf.
Power quality has become a white-hot topic in boardrooms and on Wall Street. Recent IPOs for flywheel-based uninterruptible power supply companies, such as Beacon Power and Active Power, and fuel cell makers, such as H Power and Proton Energy Systems raised hundreds of millions of dollars. Companies of all kinds are spending billions of dollars per year on technologies, ranging from batteries and flywheels to diesel engines, to assure a constant flow of power to their factories, clean rooms and data centers. But that reliability has a cost; much of the equipment consumes more electric power, which Mills says supports his contention that the Internet will increase power consumption.
“I don’t believe for a second that electric power demand is going down,” Mills said. “It hasn’t for 100 years, and it won’t go down now.”
So Who’s Right?
In the debate over power demand and the Internet, data centers are often exhibit No. 1. These facilities, also known as “server farms” or “telco hotels,” consume vast amounts of electricity. With power concentrations of 100 watts per square foot, a 10,000 square foot data center can demand as much power as 1,000 homes. But unlike homeowners who turn their lights off when they leave for vacation, data centers require full power 24/7. In Seattle, a raft of new data centers is forcing the city to scramble to meet their needs. Over the next 24 months, the city’s utility expects a handful of data centers to raise its average daily demand by about 250 megawatts, an increase of nearly 25 percent over current loads. Other regions, including the San Francisco Bay and Chicago areas, are also facing power supply problems caused, in part, by data centers.
There is little doubt that electricity usage is rising. Until recently, the EIA was projecting that domestic power demand would rise at a rate of 1.3 percent through 2020. But in late November, the agency increased its forecast by 38 percent, to an annual growth rate of 1.8 percent, citing higher than projected economic growth and a “reevaluation of the potential for growth in electricity use for a variety of residential and commercial appliances and equipment, including personal computers.”
Although the EIA listed computers as a possible reason for the increase, economists and energy analysts cannot say with certainty why electricity usage is going up, how fast it will grow or even the best way to meet that new demand. Nor is it clear how much of that growth is being caused by Internet-related facilities like data centers. Some experts, like Taub, believe that much of the growth is being caused by the “wealth effect.” Americans are making more money so they are buying more gadgets that use electricity. In addition, people are buying bigger houses that require more lighting, air conditioning and other comforts that require lots of power. And as the American economy grows, more companies are launched, more stores and offices are built, and thus, more electricity is consumed.
Mills agrees that the wealth effect is playing a role, but he firmly believes data centers will increase consumption. “When we get tens of millions of square feet of data centers, the power supply problem is going to be more acute,” Mills predicted. In fact, the two sides find some common ground when it comes to power consumption by data centers. Jonathan Koomey, a staff scientist at Lawrence Berkeley Labs, has calculated that by 2005, data centers could be consuming slightly more than 1 percent of all electricity in the U.S. Koomey based his projection on a report written earlier this year by Richard Juarez, a senior Internet analyst at Robertson Stephens, who predicted that 50 million square feet of data centers will built.
Juarez’s estimate may be conservative. On Dec. 12, IBM chairman and CEO Lou Gerstner announced that Big Blue will build 50 new data centers to meet the growing demand for the outsourcing of information technology services.
Still, Koomey and energy efficiency experts such as Amory Lovins, the CEO for research at the Rocky Mountain Institute, an energy think tank, believe that Huber and Mills drastically overestimated overall demand from the Internet. For one thing, they say, the two erred in estimating that each computer now on the Internet uses 1,000 watts of power. In reality, the average desktop unit and monitor use about 150 watts. When in sleep mode, they say, that figure drops to 50 watts or less. Laptops are even more efficient, with some newer models using less than 30 watts.
Furthermore, they charge that Huber and Mills are carrying coal for the mining and utility industries. Last year, a day after the Forbes article appeared, the Greening Earth Society, a group financed primarily by companies that mine, transport and burn coal, published a report authored by Mills titled “The Internet Begins with Coal.” One of Greening Earth’s primary objectives is to cast doubt on the science behind global warming theories and to promote the message that “the earth is actually getting greener thanks to increasing CO2 levels.”
Huber and Mills are the co-editors of the Huber-Mills Digital Power Report newsletter, which is published by the Gilder Technology Group, the Housatonic, Mass., company headed by George Gilder. In their Forbes article, Huber and Mills predicted that the “infoelectric convergence” would result in a massive increase in energy use. Moving two megabytes of data on the Net, they said, requires the energy equivalent of one pound of coal. And with hundreds of millions of new digital devices, ranging from digital X-ray machines to Palm handheld computers, getting plugged in, the future of our economy depends on burning more fossil fuels, including coal, which produces 56 percent of the electricity used in America. All those Net-related electronics consume “up to 290 billion kWh [kilowatt hours] of demand. That’s about 8 percent of total U.S. demand,” they wrote. “Add in the electric power used to build and operate stand-alone (unnetworked) chips and computers, and the total jumps to about 13 percent. It’s now reasonable to project that half of the electric grid will be powering the digital-Internet economy within the next decade.”
Several statistics counter those assumptions. For instance, Koomey and Alan Sanstad, a staff scientist at Lawrence Berkeley, point to the EIA’s 1999 calculation that all personal computers, including residential and commercial, use just 1.4 percent of the nation’s power. In fact, EIA data reveal that American homeowners use twice as much electricity running color televisions than they do operating PCs and that TV power consumption will grow at a faster rate than demand from PCs will grow. (An EIA stat that supports Mills predicts that electricity use by PCs in the commercial sector will grow at 5.1 percent per year through 2020, a rate far above that predicted for overall electric consumption).
Critics and Politics
The sharpest criticism of Mills has come from people like Lovins, who calls Mills a “liar” and Joe Romm, the executive director of the Center for Energy and Climate Solutions, who has written extensively about how the Net is increasing the efficiency of the U.S. economy. Last year, a report he co-authored about the Internet and global warming argued that computers and the Internet were creating structural changes in the economy that were reducing the demand for energy. EIA data appear to support his thesis. The amount of energy needed to produce $1 of gross domestic product has been falling since 1991, and that trend has accelerated dramatically. Since 1996, energy intensity has fallen at an average rate of 3.2 percent per year, a phenomenon Romm says is difficult to explain without including the effect of the Internet.
Lovins and others explain that the equipment used to power the Internet is becoming more efficient, thanks in part to the Environmental Protection Agency’s Energy Star program, which is reducing the power demands of personal computers. What’s more, they say, the basic building blocks of the Internet are also becoming more efficient. Intel’s Pentium-class processors can require several dozen watts of power. Lovins points out that the new Crusoe processors from Transmeta, which will compete with Intel’s chips, “draw at most six watts.” And it won’t be long, he predicts, before manufacturers begin building more energy efficient servers to reduce their operating cost.
Despite Lovins’ projections, there’s a nagging question: If the Internet is making the economy more energy efficient and the 8 percent figure is wrong, why have Huber and Mills received so much attention while Lovins and his allies have largely been ignored? Romm accuses Huber and Mills of making radical estimates to draw attention for their $295-per-year newsletter. (Full disclosure: this reporter is a subscriber). “You get media coverage if you adopt an extreme, off-beat, unusual point of view,” Romm said. Others, like Taub, blame it on the “Gilder Effect.” Since Huber and Mills are associated with Gilder, many people assume they must be correct. “And if you repeat it enough, it starts to gain credibility because you hear it from so many sources,” Taub said.
Mills says his critics are missing the point; he no longer cares what percentage of domestic electricity the Internet consumes. “I have chosen not to engage in the question of whether it’s 5 or 6 or 8 percent,” he says. Lovins and Romm can discuss global warming all they want, says Mills. They want to “save the planet,” he said. “I’m not interested in saving the planet….My objective is to keep silicon lit. And there’s going to be a huge market for people who want to keep silicon lit.”
Wall Street is still skeptical. It’s too early to call this,” says Sam Brothwell, a stock analyst at Merrill Lynch who has been monitoring the debate. “I think there’s anecdotal evidence that computer usage is growing and it’s having a significant impact” on total electricity consumption, he said. “But it’s very hard to translate that into a number.”
Right or wrong, Huber and Mills have left their mark on discussions about power and the Internet. And given the shortage of power in California and the EIA’s projections for increased electricity usage, the utility industry is investing billions of dollars in new power plants. Lovins hopes that much of new demand will be met by nonpolluting technologies, such as fuel cells, and by renewable energy sources. “Wind power is growing 25 to 26 percent per year,” he says. And photovoltaic cells, he predicts, will soon compete in price with other generation methods.
Mills says he, too, favors using fuel cells and renewable energy sources. But, he doubts they can supply all the power needed. “We have to face real life. We have to provide power for the Internet economy” he says. “And that means burning gas, oil and coal.”