My computer has a mass of about 1.7 tonnes. So does yours. It probably doesn't feel that way, but that's the mass of the materials that went into its making.

You'll probably not be surprised to learn that most of it is water: 1.5 tonnes for a typical desktop PC with a 17-inch monitor. Most of the rest is 'embodied energy': producing this equipment requires around 240kg of fossil fuel.

These are the figures cited by Eric Williams of the United Nations University, Japan, in Computers and the Environment: Understanding and Managing their Impacts (Kluwer Academic & UNU, 2003). The question is: should we worry? It's again no surprise that the manufacture of high-tech electronic devices is energy-intensive, or that the economy in materials that apparently stems from miniaturization masks a vast investment in raw materials and energy needed to build at ever-smaller scales. Thermodynamics alone tells us that we do not get such intricate organization of matter for free.

But the question is of course too simply stated. High-value-added, lightweight materials may be costly in all respects to manufacture, but they might then significantly reduce the fuel consumption of a vehicle. And what about the relative environmental impacts (and the associated costs) of different materials during manufacture or after disposal?

What's more, technical advances do seem capable of reducing the total costs of technologies: a study from 1993, based on data from the late 1980s, estimated the embodied fossil-fuel energy in a computer workstation at around 740kg, three times the figure calculated by Williams. Improvements in processing efficiency over the past decade may account for the difference.

Perhaps the value of such studies lies primarily in forcing us to think more carefully about how we use these devices, as well as where to focus efforts to reduce their overall cost to the environment. Making a car requires around two tonnes of fossil fuels: eight times more than a computer, but nearly an order of magnitude less on a weight-to-weight basis. Perhaps even more pertinently, the typical active lifespan of a computer is a fraction of a car's. We change our car about once every ten years, but might buy a new computer every two years (and who wants the old one?).

Then there is the question of energy use over a lifespan. For a refrigerator, 96 percent of the fossil fuel consumed in its life cycle is burnt up during use; for a computer, three quarters of the total energy goes into making it, and only a quarter into powering it.

The answer, then, seems obvious: computers, or at least their parts, should be reused whenever possible. Their components, particularly the microchips (94 kg of embodied fossil fuel in each computer), represent an enormous energy investment. But somehow the economics have gone awry. Why go to the expense and trouble of upgrading your tawdry old model, which doesn't have the compatibility you need anyway? The first obsolescence-proof computer will spark a revolution.