Intel wireless act a risky high wire endeavour

UNTIL THIS YEAR, Intel did not make radio chips. Now, it is not only planning to make them for Centrino, but to develop a whole new class of digital radios based on silicon, which would be small enough to be incorporated into all devices and could combine Wi-Fi, Bluetooth and cellular (and their successors) in one low cost system.

This is only the latest in a whole series of plans designed to make Intel as dominant in wireless and mobile as it is in PCs. But is even a company of Intel's scale biting off too much this time? It is competing on an awful lot of fronts with the specialists, and already Centrino has come in for some criticism for delivering lower performance than its rivals. Better to combine Intel's Pentium with wireless chipsets from the experts, goes the line. And by deciding to make everything at home, Intel is alienating all the third party chipset makers with whom it has had an understanding - if sometimes an uneasy one - all these years.

Then Intel is moving into cellphone chips as well with the XScale architecture. Its strategy for operator and developer alliances is sound, but now Intel is taking on a whole new range of powerful rivals, opening up a new battle front, as if competing for the high end server with Itanium were not enough. Not enough to challenge IBM, let's take on Texas Instruments and Motorola in cellphones and Intersil, Cisco and Broadcom in WLans too. None of these companies will surrender ground happily or easily.

And now Intel is even planning to make the really specialist chips, such as radios, at the same time that it is spending $150m of its venture capital fund and almost half a billion of its R&D budget on wireless.

The radio plan is not just part of a plan to stop relying on third parties - it has a far larger and longer term goal, rather like its development of a personal data server, which would allow users to carry data an applications on a cellphone-like device and access it wirelessly from any PC that happened to be within range. Intel wants to change the whole shape of personal computing to ensure that it remains in control even as the traditional Wintel model starts on its slow decline.

Key behind this ambition is Intel's first ever CTO, Pat Gelsinger, who believes that if wireless is going to be everywhere down to the wristwatch, Intel needs to be in all those devices too. He outlined his big idea under the slogan ‘Radio Free Intel' earlier this year, but has now unveiled preliminary results of work on silicon radios that could reset the goalposts in mobile devices.

The results were presented at an international semiconductor symposium in Japan this month and have gone a long way to mending a rift within Intel over Gelsinger's Radio Free pronouncements. One faction in the company, apparently including CEO Craig Barrett, was concerned that Gelsinger was promising something beyond Intel's capabilities to deliver. Despite a new confidence springing from the early demos - though the tiny silicon radios will not be commercialized until 2005 - Barrett may still be proved right.

Intel's development would transform the wireless radio from an analog technology to a digital one, a huge technological leap to convert gigahertz waves into digital representations. The timetable is to commercialize these breakthroughs into very low cost, multi-network chips by 2005. By 2010, Gelsinger sees Intel offering ‘agile radios', which will not only pick up different types of connections, but will seamlessly and intelligently switch between them for optimal performance.

Intel may be spending a hefty chunk of its $4bn R&D budget on this, but the stakes are high too - a silicon radio would promote its overall ambition of making its chips ubiquitous in all forms of wireless devices; and would have the very valuable side-effect of prolonging the return on Intel's huge investment in Cmos processes and capacity. But Motorola and Philips are both said to be looking at similar developments, and Intel itself admits there are high technological hurdles to cross in the coming year. The strategy does serve to illustrate once more, however, Intel's determination to utterly control a wireless market that has not been its traditional home, and its willingness to take high risks and high technical leaps to achieve that and to leapfrog the incumbent giants.

In the nearer future, the latest market figures on WLan chips show how Intel is creeping up behind the established leaders. Both the top two positions in the Wi-Fi chip market will be usurped by next year, with Broadcom and Intel fighting for the top spot at the expense of Intersil and Atheros.

The WLan chipset market has become crowded on the back of the Wi-Fi boom, and now numbers a dozen serious suppliers. Intersil is the leader, with Broadcom poised to take second place by the end of this year, overtaking Atheros. Between them, these two will control about half of the market. The other majors are Agere, Atmel, Marvell, RF Micro and Texas Instruments. Now Intel and AMD are throwing their hats in the ring, in Intel's case with potentially highly disruptive effect. So far it offers modules rather than a complete chipset - it has a MAC but uses radios from Philips, for example. But Intel plans to make all the Wi-Fi elements of its Centrino family itself, and has a dual-mode a/b chipset, made entirely with Intel silicon, ready to launch later this year. Its production capacities and the heavy weight that Intel is putting on Centrino make it a powerful disruptive force in the formerly cosy world of the Wi-Fi chipmakers.

Radios, Wi-Fi chipsets, cellphone silicon, PDA processors and personal data servers - and that's just Intel's wireless and mobile effort, currently lossmaking but with the potential to be more important than Pentium and Itanium put together. But can a company that has been used for years to enjoying a near-monopoly, with just AMD there nipping at its heels and keeping the antitrust authorities away, really adapt to challenging for the most important market in computing, against about 30 rival vendors with serious technology and established relationships? It never does to underestimate Intel, but this time it may be showing the rashness of the megalomaniac. µ