News

Taking on the mobile giants

For a start-up company to take on the giants of the industry for its first product seems like an insanely ambitious strategy, but three year-old Bristol-based Icera Semiconductor believes it can take on, and beat, the biggest companies in the world making mobile baseband chips.

'We started off with a passion to be a big European-based chip company, a company which could go public, a company that was sustainable,' says Stan Boland, founder and CEO of Icera. 'The biggest and most challenging chip market is wireless. Particularly cellular. Cellular is difficult. There are big barriers to entry. We decided that if we wanted a long-term sustainable business we would have to own the baseband technology. If you don’t own the baseband, you’re not in control of the platform. So baseband is absolutely the place to be.'

But the $6bn mobile baseband market is dominated by some of the biggest companies in the semiconductor industry such as, Texas Instruments, STMicroelectronics, Freescale Semiconductor, Qualcomm and Philips Semiconductors. Is it rather ambitious for a start-up company to take on these industry giants?

'We have respect for the engineers in our competitors, but we know we are at least as good. There are no barriers in our mindset, ultimately it comes down to the raw talent we can assemble,' says Boland. 'We have built a great, great team. There’s no question it’s the strongest ever chip team assembled in Europe, if not the world. We can propel ourselves from no position to the leading position at our entry point into the market which is HSDPA.' HSDPA is an enhancement of 3G mobile technology which boosts downlink data rates.

Boland’s confidence stems from his previous success as CEO of Element 14, a DSL start-up, which was sold to Broadcom for $600m in 2000. 'The last technology cycle was largely about infrastructure. This technology boom, which I think will be more long term, is based on the consumer,' says Boland. “We think this allows Europe to build a large public company like Broadcom and Qualcomm, except pure-play wireless and open.'

So Icera’s aim is to dominate the mobile handset platform by out-engineering the industry’s giants on the baseband processor, and then leveraging its success in the handset to build a major wireless-based European semiconductor company. That is pretty ambitious.

'We have the confidence that we in Europe can do it. The chips developed by Element 14 went on to be market leader,' says Boland. 'We have great engineers who are able to look at problems in a different way to how competitors look at them. There is an ability in Europe to create great processor applications. We think about how to solve the maths problem in these algorithms, and can remove the constraints and push all that complexity into software.'

Icera’s engineers have created a completely soft modem which means it can use a small chip with low power consumption. Software constantly adjusts the algorithms to the channel conditions. If that was done in hardware, it would need a large chip consuming a lot of power.

'The reality is that our solution uses significantly less power than any competitor’s modem solution,' says Boland. 'When we did Element 14, our chips were smaller and used less power than any of the competition’s chips. Those were used in infrastructure. Those advantages make a bigger difference in the handset. Our chip gives a big saving in the die area and a big saving in power.”

At the full HSDPA data rate of 3.6Mbit/s, Icera’s stacked die package is thought to dissipate well under 0.5W. One of the stacked die is a 90nm CMOS digital die for processing algorithms, the other is a 0.18µm analogue baseband chip. The two chips’ 324-ball BGA package measures 12x12mm. Icera’s name for it is Livanto.

'We can design a lower cost handset by doing everything in software,' says Boland. 'We can implement much more adaptive algorithms, while doubling and trebling the data throughput, giving the operators more voice traffic and so they can support more users.'

Most of the cost and effort in implementing HSPDA is the signal processing in the handset. The standard uses adaptive techniques to modify the code rate, modulation scheme, the number of codes used and power per code. A hybrid ARQ mechanism is used to efficiently re-transmit data.

'Our opportunity is that the number of different air interfaces is increasing,' adds Boland. 'One of the benefits of using our architecture is that we can support multiple air interfaces on the same silicon. It just requires more programs that all go on the same silicon.'

If he wanted to found a significant European chip company, why did Boland sell Element 14 to Broadcom rather than going for an IPO and growing Element 14 into a major company?

'Element 14 we did at a time of a financial boom. We had people beating a path to our door offering inflated prices,' replies Boland. 'They offered a 32 times return to the investors. It would have been mad to turn it down. And we thought we can always do another company.'

Why did he choose Bristol for Icera when Element 14 was a Cambridge-based company? 'We needed wireless algorithm skills and full custom processor skills. Steve Allpress heads up algorithms, and he was rooted in Bristol and he felt we could recruit some of the staff in Bristol,' replies Boland. 'And Bristol is the only centre in Europe where we could assemble a crack team of full custom processor people. That is the legacy of Inmos. They’re people who are either at STMicroelectronics or Inmos, some went through Element 14, some at DEC, some at Intel.'

It is not just Boland and his team who are confident that they can out-engineer the giants and dominate the mobile baseband market. Venture capitalists Accel Partners, Atlas Venture, Benchmark Capital and Amadeus have backed the company with $82.5m.

Icera’s latest cash injection of $40m in March will see it through until the company becomes cash generative. 'We are currently on a $2.5m a month burn rate,' says Boland. 'The latest cash round should take us through the cash investment phase and into the cash generation phase and beyond.'

At the present rate of cash burn, $40m will support another 16 months of running costs which will see Icera through until at least the summer of 2008. Long before that, Icera’s chips will be in end-products being sold to consumers. An IPO is pencilled in for 2008.

'Later on this year we’ll be seeing some products out on the market using our chips. The initial sales will be in data cards to customers, which is a much bigger market than it has traditionally been in the past,' says Boland. 'This year maybe five million data cards will be sold. It’s not really a phone but it requires the baseband technology and it is quite easy for us to develop. It is one line of business we will be supporting.'

The big market, of course, is handsets. For this a complete engineering system is needed with hardware, plus the protocols, plus the platform. Icera reckons that by 2008 the majority of 3G phones will have HSDPA.

'We need to develop a compelling platform for HSDPA from which the handset manufacturers can build,' says Boland.'All the major chip manufacturers are developing a platform. A complete platform has to be the basis on which manufacturers build the phone. We’re not fully responsible for all the applications a particular platform can offer, the challenge is to co-develop, and support, and sell it.'