Want to help cancer or AIDS research? You can give money to an appropriate charity - or if you have a personal computer you can donate its number-crunching power to the cause.

As scientists are beginning to realise, the estimated 250 million PCs worldwide linked to the Internet represent a distributed computing source that is far more powerful than any individual supercomputer.

The latest effort to tap it will be a collaboration to discover cancer drugs, involving Oxford University, the Washington-based National Foundation for Cancer Research, and two US companies, United Devices and Intel.

Graham Richards, Oxford's head of chemistry, expects hundreds of thousands of PC owners to participate in what would be the world's largest computational chemistry project.

The molecular modelling software was developed in Oxford and the distributed computing expertise comes from United Devices, a company set up in Texas to commercialise the technology behind the pioneering Seti@home project.

Seti@home demonstrated how the spare capacity of PCs could be harnessed for science - in this case, the search for extraterrestrial intelligence. Since the project started at the University of California, Berkeley, in 1999, 2.8 million people have given time on their PCs to scrutinise radio-telescope signals for anything that could come from an alien civilisation.

According to David Anderson, chief technologist at United Devices and director of Seti@home, its computing rate is twice that of IBM's ASCII White, the world's fastest supercomputer. Although Seti@home has been a tremendous success with its technology and in capturing public imagination, it has not yet detected any signals from ET.

Drug discovery projects offer participants a better chance of a hit. The first was FightAids@home, a partnership between the Scripps Institute and Entropia, a competitor of United Devices based in San Diego. It has enlisted 18,000 PCs so far to test millions of candidate AIDS drugs against computer models of HIV, the virus that causes the disease.

These distributed projects rely on the Internet to send packages of work out to participants and then back to the centre when they are finished. The PC software acts like an intelligent screensaver, exploiting the processor's idle moments without degrading its performance. It switches off instantly as soon as the user needs the power for another task.

United Devices and Expedia are planning further "public interest" research projects, for example, in environmental science and climate change. These will be combined with money-making contracts for commercial customers, for example, to design and develop complex products.

They hope to draw members into their networks both by appealing to their altruism - helping to fight disease or global warming - and by offering financial inducements. For example, United Devices runs sweepstakes with substantial cash prizes.

Distributed computing is the high-performance end of the wider phenomenon known as peer-to-peer or P2P computing, which has been made a household word by Napster, the popular music file-sharing service.

Although applications involving home computers have inevitably caught the public attention, there could be just as much scope for companies to use their own computers. Many have large numbers of PCs and workstations connected by high-speed internal networks, which are largely unused outside the working day.

"If our project is successful, it could change the way the pharmaceutic industry carries out research," says Prof Richards. "Glaxo and Merck must have many thousands of PCs inside their firewalls, which they could use for distributed computing."

Intel, the chip manufacturer, has been an important cheerleader for P2P from the start. In one view, this seems counterproductive - by enabling companies to make more efficient use of their existing computing resources, it could reduce demand for new equipment, which would hit Intel's microprocessor sales.

But Intel sees it differently. "We believe the increase in computer-intensive applications will spur the need for more computing resources," says Les Vadasz, executive vice-president at Intel.

Mr Vadasz is backing this belief with money. As head of Intel's venture capital group, the largest corporate venture capital fund in Silicon Valley, he manages a portfolio that holds investments in more than 600 companies.

"We are currently evaluating more than 100 business plans of peer-to-peer companies," he says. While many of these start-ups have promising technologies, he notes that building a viable business around P2P is not easy.

Other large companies are also investing in P2P start-ups. GoldPocket Interactive recently received a third round of funding from AOL Time Warner Investments. It did not disclose the amount but it adds to $55 million raised in earlier rounds for its interactive television service, which allows millions of users to interact with each other and with TV programming through a P2P model.

While many companies see P2P as the next big thing for the Internet, there are many concerns. IT managers are most worried about security. This is less of a problem for internal networks but when several networks from different sites or companies are to be combined, security is paramount. Even the non-commercial Seti@home has been attacked by hackers trying to simulate alien signals.

There is also the issue that individuals might switch off their computers and disrupt a background P2P process at the wrong moment. And the cost of managing IT resources will rise if thousands of computers are roped together to run very intensive applications. Systems management is one of the highest costs in any IT department and it might outweigh the savings on hardware made by more efficient use of computing resources.

Even the hardware savings might evaporate. Companies need to upgrade their systems to handle the additional burden of P2P applications - good news for Intel and PC makers. A cost-benefit analysis might come down in favour of more conventional supercomputers, which are becoming steadily cheaper and pose less of a security risk.

For the "enterprise applications" that form the core of most companies' IT systems, there are still no compelling applications - and no easy way to split those tasks across thousands of processors.

However, science rather than commerce has led to most of the historic changes in IT practice, from the first electronic computer to the worldwide web. Distributed computing may be the next step to be promoted by science.