Michel Mayer, chairman and CEO of Freescale Semiconductor, is a man with a mission. The mission is to turn round what was always seen as a difficult business - the old Motorola semiconductor division, now renamed Freescale Semiconductor and spun-off as a separate, publicly quoted company.
“People say it takes seven years to change the culture. Behaviour changes quickly, but the underlying culture doesn’t,” says Mayer. “But in three years time we will be an established, recognised leader in the industry. We’ll get there.”
Could that target be affected by an intervening industry downturn? “There is potential for a correction but not as bad as 2000,” replies Mayer. “The growth has not been as steep as it was in 2000. The supply chain is much more agile now.”
Is he prepared to sacrifice profitability for growth? “I won’t get into businesses where I can grow revenue, but where I know that it will be a low margin business,” responds Mayer. “But will I invest and accept low margins while getting established in a business? Yes, I will. But I don’t see value in trying to be the biggest semiconductor company in the world. Our objective is to be a leader in the spaces we choose to serve. Leader doesn’t mean being number one. It means being number one, two or, in some segments, number three.”
Freescale has dramatically increased the number of patents it is being granted. Why is this? “We sat down and decided we were going to double our rate of patent filing,” replies Mayer. “We introduced incentives, we paid people more for filing patents, we increased the numbers in the legal department.”
Mayer says one of the reasons for doing this is to have a revenue stream from licensing. But the main reason is “to maintain our IP position. It’s important to have a strong IP portfolio. It gives you a good defensive position”.
With fabless competitors like Broadcom and Qualcomm, why does he continue to run and invest in fabs? “The primary driver for having fabs is to supply the automotive markets,” replies Mayer. “Automotive requires analogue ICs and sensors which are specialised technologies. You can’t go to TSMC and get sensors built.”
He believes that Qualcomm does not have to be cost competitive on chip manufacturing “because of their strong IP position. They are priced for value, like Windows XP”.
“There are spaces where you need fab and spaces where it adds value. Can you do it without fabs? Yes. But if you look at our return on capital and a return on sales they’re not bad. And, in the long run, the few remaining IDMs will have an advantage,” he concludes.
As the CEO of a major wireless technology company, Mayer sees some of the failings of the network operators as frustrating. “The reason for the slow roll-out of 3G is because of quality of service issues, there are a lot of announced roll-outs that are partial,” reckons Mayer. “There is a long way before every basestation is upgraded to 3G. As a user I am amazed that they get by with such a poor quality of service. It is difficult to imagine video to the phone when so many voice calls lose their connection.”
Mayer claims one of the reasons for the success of the Blackberry is because it is reliable. “There’ve been so many times I have to send Blackberry messages saying ‘I’m sending this because I can’t reach you on the phone’. My personal view is that consumers should have forced the industry to make the capital investment to improve the network.”
He is also not afraid of commodity markets which are so often seen in the semiconductor business as unprofitable.
“In peoples’ minds there is a clear connection between low cost and low margin. The reality of the semiconductor industry’s economics is that people who make commodities sometimes have better margins than people who make these galactic hugely complex ICs,” says Mayer. “For instance Bluetooth is a commodity. It’s a $2 part. There are two leaders in Bluetooth. Look at CSR’s growth margin. They can have 50 per cent gross margins on a 20 cent piece.”
According to Mayer the competitive landscape is what is more important with margin. “Competition drives margins down, for instance look at Intel and AMD. But you can find spaces where there is a niche and not much competition. Maxim and Linear are making a fortune on little analogue parts that are very cheap. Look at the margins on flash. Is that a commodity or not? Look at the margins of Samsung.”
One of Freescale’s long-term technology commitments has been to MRAM, but can it succeed?
“MRAM has the potential to be widely used, it has to go down the cost curve,” replies Mayer. “Today it is niche. There is nothing intrinsically in the technology that says it can’t go down the cost curve. I don’t know of any showstopper why it can’t be widely used.”
But Cypress Semiconductor pulled out of MRAM last year, with CEO, T. J. Rogers, saying that the technology could not be made cost competitive with SRAM. “Ours works,” responds Mayer.
Freescale’s MRAM cell is four times smaller than a six transistor SRAM because its cell uses one transistor and one magnetic tunnel junction cell (MTJ) and the transistor sits on top of the MTJ.
With both read and write speeds of 35ns, the Freescale MRAM has significant potential as a replacement for mid-range performance SRAM, but it is not fast enough to compete with high-speed SRAM which has sub-10ns access times.
The firm is positioning and pricing MRAM to be a replacement for battery-backed SRAM. However, with a cell size four times smaller than SRAM, the Freescale MRAM has the potential to get considerably denser than the current 4Mbit which is made on a 0.18 micron process. For instance a 90nm process delivers a 64Mbit MRAM.
However, as now constituted, Freescale’s MRAM cell will always be bigger than a NAND flash cell on the same geometry and so MRAM will not compete with NAND flash.
Freescale’s intentions for MRAM lie mainly in the embedded space, principally because embedded MRAM requires only six extra masking steps on top of standard logic.
The key part for the firm is the Power PC, now renamed simply Power, which Freescale is actively seeking to proliferate as a ubiquitous and global microprocessor architecture.
“Licensing Power cores will get easier and easier to the point where we might start giving them away to universities and things like that,” says Mayer. “The goal is not to go head-to-head with ARM and compete in cellphones.”
Because the Power microprocessor has so many applications the common architecture evolves along different development streams, for example, graphics, communications, servers and computing, printing and imaging, automotive, networking and industrial/automotive.
So a Power die can vary in size from 1mm2 to 30mm2, depending on the application, but they all have binary compatibility at the user level, and all run the same instruction set.
Mayer’s mission is clear: to be an industry leader by the end of the decade. His message is: collaborate and engage, with customers, with rivals, with academia and with the electronics community in its largest sense.