For an engineer, John Roth has an unusual hobby: he makes stained-glass windows. "I used to know he was on a creative tear because he'd come in with his fingers all bandaged up," says Tom Hennebury, a long-time friend and associate.

Mr Roth took the helm at Northern Telecom, the Canadian company which is one of the biggest in the world telecommunications equipment industry, last October. A 29year company veteran, he has taken up the helm at the company's gleaming new headquarters in Brampton, a Toronto suburb.

Less than a year after becoming chief executive of Nortel, he launched what is probably the biggest project in the 103-year history of the company which employs 900 people in Galway. Nortel announced in June it was acquiring Bay Networks, a San Francisco data networking company, in a deal worth about $7 billion (£4.7 billion) when it was concluded last month.

The acquisition is intended to catapult Nortel, which became a global force on the strength of voice switching, into the rapidly expanding business of installing networks to carry data traffic. That market, driven by Internet demand, is growing 10 times faster than voice traffic.

"We're looking at what's next. And the Internet is what's next," says Mr Roth. "The more we understand it, we can see the need to overbuild, rebuild or bolster today's networks in a way that will carry that type of traffic much better. And we need to become a strong player very quickly in that segment."

Nortel's goal is to be the first company to develop what Mr Roth has dubbed "Webtone", in which customers can access the Internet with the same speed and reliability they expect when making telephone calls. Web pages, teleconferences, videos, home shopping and online banking will all be instantly accessible.

While about 75 per cent of North American Internet traffic already travels over Nortel's fibre-optic networks, those networks were built for voice rather than data traffic. And Nortel, at least prior to the Bay acquisition, was not a leader in designing networks specifically for data.

So far, investors seem to be far from convinced by the company's new direction. Nortel's shares dropped 15 per cent after the Bay deal was announced and have continued to fall. The company's announcement earlier this month that it would shed 3,500 of its 80,000 employees worldwide, including closing two British plants, did not restore confidence. The company has said its Irish employees will not be affected as the Galway plant has been working on a separate plan to streamline its business.

Nortel stock is trading near 60 Canadian dollars (£26.50) down more than 40 per cent from its 52-week high.

The decision to buy its way into the data market was a radical departure for Nortel. The company became Canada's largest and most successful manufacturer by developing its own new technologies a little faster than its competitors.

Mr Roth seems an unlikely candidate to break with that tradition. He began his Nortel career as a 27-year-old engineer in the company's Ottawa research lab, and has seen the handsome profits generated by technological breakthroughs.

In the 1970s, Nortel researchers pioneered the digital central office switch, which transformed Nortel from a purely Canadian company to a leading force in US and world markets. Last year, 54 per cent of its $15.5 billion in revenues came from the US, 22 per cent from Europe, 15 per cent from Asia and just 9 per cent from Canada.

But Mr Roth has a theory about how the next breakthrough will come, and he says Nortel cannot do it alone. It will happen by developing internal data networks for corporate customers that are then refined to meet the rigorous demands of the public carriers, Nortel's biggest buyers.

The digital central office switch, he says, was developed by selling smaller business communications systems, known as private branch exchange or PBX, to corporate customers.

"Corporate customers drove us to create the features and functions they needed to improve their productivity," he says.

That led to improvements which made those technologies sufficiently robust to meet the requirements of public carriers serving millions of customers.

"We predict the same thing's going to happen here," he says. By developing data networks for corporations, Nortel will be able to design similar systems for public carriers.

Enter Bay Networks. Mr Roth spent the best part of a year talking to US data networking companies like Cisco Systems, Ascend, 3Com, Cabletron and Bay.

Throwing out a company like Cisco, which is probably rich enough to buy Nortel rather than the other way around, Bay was the best fit, he says.

Bay has what he calls the world's most reliable router, the key technology for directing data traffic. More importantly, its customers include the most demanding corporations such as banks, airlines and stock exchanges.

There are fears the Bay deal may be Mr Roth's first big stumble. Bay has long been an underperformer financially; and that, rather than giving Nortel instant credibility in the data market, has saddled it with a big turnaround project, says Maribel Lopez, a consultant with Forrester Research in Boston.

Further, while the coming convergence of the telecommunications and data industries has become almost cliche, the corporate cultures of the two industries are vastly different. Critics think Mr Roth's decision to leave Bay as a largely independent arm of Nortel, with its own headquarters in California, will not make the task of integration any easier.

Mr Roth acknowledges the difficulties, but says the two companies are remarkably complementary.

Further, he argues, his traditional competitors such as Lucent, Alcatel and Siemens will have to undertake similar mergers or alliances with data companies, or risk dying slowly. And there are not very many partners to go round.

"We'd rather be the first to move than the last to move," he says. Even if it means a little blood on the furniture.