The world is often split into winners and losers. Like it or not, some companies are faced with David and Goliath type confrontations, taking on rivals who are often an order of magnitude larger than they are.
So how are they able to cope with such a daunting prospect, and what can they do to even up the odds? Are the roles fixed, or is there still the possibility of a firm with a bit of tenacity and inventiveness to rise out of the mire?
The ‘haves’ and ‘have-nots’ scenario seems to be played out to its extreme when you examine the programmable logic arena. Here the top two companies (namely Xilinx and Altera) take almost 85 per cent of the business, leaving the poor relations to fight over the scraps.
So is there hope for the little guys, or are they forever consigned to the ‘also ran’ category? Let’s face it, we all like to see the underdog win from time to time, but the chasm that separates the two acknowledged market leaders and the wannabes is still immense.
 |
|
| Wim Roelandts |
|
It’s not surprising that Wim Roelandts, CEO of market leader Xilinx, is pleased with himself: “We have grown market share every year for the last six years and we believe we will continue to do so. The question is more where the future opportunities lie.”
With the advent of the company’s ASMBL architecture, and the range of application-optimised Virtex-4 FPGAs that make use of this, it clearly has strong intentions to eat into the far larger Asic market. Clearly positioned as the big fish in this relatively small pond, the easiest strategy to employ in order to grow is to explore new areas, and the $15-20bn Asic/ASSP space seems to be the obvious place to start.
“We see lots of opportunity for us here. The best way for us to pull further away from the rest of the pack is to make sure we get more business from Asics than our competition does,” muses Roelandts.
He sees programmable logic rapidly becoming a closed shop, “There are huge barriers to prevent anyone entering this sector. The first is the speed of innovation, some companies have spent years developing a new product, then found that by the time it comes to market it is already obsolete. The second is software, there is no real independent software industry to serve this business, so you’d have to commission a hundred or so engineers for a couple of years to come up with the tools needed.”
Jordan Felburn, principal analyst at iSuppli, supports this appraisal: “It is certainly very difficult to break into the programmable logic domain. Texas Instruments, Agere, Freescale, and many others have tried and failed. The investment needed is big, and the chances of dislodging one of the incumbents are too high to justify it.”
This puts the established suppliers in an enviable position, protected from outside invasion and seemingly too powerful to be threatened by their close neighbours. Roelandts points out: “Normally when a market is growing the number of competitors increases, but here we have the unusual situation where as the sector continues to expand the players present are actually shrinking.”
 |
|
| John Daane |
|
Altera has been engaged in a back and forth battle with Xilinx for programmable logic supremacy for many years. Its CEO, John Daane, in what must be one of the few things that they agree on, echoes Roelandts sentiments: “If you look at the statistics, Xilinx and Altera have both gained market share, and everybody else has lost out.”
However, he does not think that this has not done any harm to customers. Despite this seeming duopoly, the pace at which innovation is taking place is still managing to out-run the chip industry as a whole. Performance levels continue to rise and cost per logic gate is far less that it was only a short time ago.
“The competition between the two of us is still fierce,” says Daane, “we invest in new products at an incredibly high rate, both of us putting in 19-20 per cent of our revenue, which is exceptional for the semiconductor world.” He also feels that some form of natural order has been set, though only question is whether his or Roelandts’ firm are at the top of the heap. In his opinion: “For another company to go head-to-head with us will be increasingly hard, and is certain to take a lot of outlay.”
This is not just bravado on Xilinx or Altera’s part, the figures back it up, it has to be said that the programmable logic market appears to increasingly look like a two horse race, and the gap between them and the chasing pack is still widening.
Total FPGA/CPLD Market Share 2004
| Company |
2004 |
2003 |
| Xilinx |
51.5% |
49.0% |
| Altera |
32.3% |
31.8% |
| Lattice |
7.8% |
8.5% |
| Actel |
5.4% |
5.8% |
| Cypress |
1.7% |
2.0% |
| Quicklogic |
1.4% |
1.6% |
| Atmel |
0.5% |
0.4% |
[Figure courtesy of Gartner Group]
Note: These figures do not include Altera’s HardCopy structured Asic products.
Based on this, it all looks pretty grim for those trying to play catch up. “There’s no doubt that Xilinx and Altera can call upon much greater R&D budgets, and are therefore able to push the performance envelope far harder,” says Bryan Lewis, vice-president and chief analyst for Asics and programmable logic at Gartner Group.
Lattice and Actel are just about managing to keep hold of the current business, while Atmel just appears to be milking any residual trade it can get hold of here, and Cypress is concentrating all efforts on its PSoC products. Lewis feels that “this means the smaller players need to find something to latch on to that will allow them to differentiate themselves”.
Actel currently ranks fourth in the pecking order, and CEO John East appears to have no delusions about the fact that their size puts them at a disadvantage. With the need for innovation being so prevalent in this sector, the market leaders can obviously put a lot more money into product development. “I think if you try to go up against these companies head on, you are going to have a tough fight on your hands,” says East, “but we aren’t trying to do that, we are choosing our battles. Actel does not have to match them dollar for dollar on R&D when they are trying to play in every market across the board, we just have to make sure that we are up with them in the areas where we are trying to take market share.”
 |
|
| John East |
|
Actel is hoping to use its third-generation ProASIC products to take a piece of the $3 or less per unit market. “We are very capable of competing here,” East believes, so as far as he is concerned, “Xilinx and Altera can slug it out for the $20 devices, but that’s not for us.”
In principle, the move to smaller architectures should favour the anti-fuse technology that forms the basis of much of Actel’s product portfolio, allowing it to move out of the purely mission critical sphere in which it has previously been confined. But although soft errors (caused by cosmic rays) will become an increasing problem, as its SRAM-based competitors move to 90 or 65nm, there is still a question mark as to whether the cost of its devices will be more competitive than just building redundancy into standard FPGA systems?
“In some space applications that is what will happen,” East responds, “but it’s not just the cost of the chips, it is also the complexity of the system, collating the data, and then deciding what is correct and what is not. In nearly all scenarios, it would just make no sense to go to this much trouble.”
The company is increasingly making use of flash-based technology in its devices. “Now we have the re-programmability aspect covered, and this has helped us to double our number of design wins. It means we can get into places that we couldn’t before,” East states.
However, with the big guys both using 90nm architectures, while Actel languishes at the 130nm mark, it has to be argued that they must have certain advantages on power consumption and chip size. But again East is eager to point out that this is only part of the story: “We think that as you get to 90 or 65nm, the issue is how much are you going to gain from each architecture shrink,” he says.
“Sure, the dynamic power is going down, but your leakage current is shooting upwards, and therefore the static power situation will get worse not better. So although they have the advantage over us when you talk about dynamic power, when it comes to static or boot up power we kill them. The way things are no, one customer worries about the dynamic power problem, and another worries about static power, it just depends on the design.
"Another issue is these devices don’t have room for a boot up memory. Now not all designs need this, but when you do it means you will have to add addition complexity to the system, because you can’t integrate things onto the chip.”
In his opinion his rivals are putting forward a false economy, he feels there is no sense in going to lower architectures if it just means your shifting your problems somewhere else on the board, the complete system will not be any easier to create that way.
Another prevalent issue is that of increased IP security. Here again Actel believes that using flash can offer something which delivers real value to the client base. With all players now looking to broaden their horizons and find new opportunities in the consumer space, the advantages of this technology over the SRAM-based devices favoured by their counterparts will become increasingly apparent.
The issues of reverse engineering mean that the standard FPGAs are much more exposed to the possibility of being ripped off, and providing Actel can keep the chips it manufactures cost competitive, then this added security is likely to make its products a more compelling option.
One final ace up its sleeve could be the ‘Fusion’ architecture, which the firm announced only this week. Here the company is setting its sight on developing the Prgrammable System Chip (PSC) market, using its expertise in flash to meet the needs of designs that would normally have to be dealt with by a mixed signal Asic.
 |
|
| Cyrus Tsui |
|
Lattice Semiconductor is also optimistic that after several years in the doldrums the tide has finally started to turn. Its CEO, Cyrus Tsui, hopes that the collaboration with Fujitsu will be a huge plus-point: “The expertise that has been placed at our disposal, both in Asics and in non-volatile, means we are now capable of fighting above our weight. The process technology is the vehicle with which we bring our product offerings to market, and partnering with a firm like this gives us access to the cutting edge that even our largest competitors will envy.”
Brian Lewis thinks that Tsui has good reason to be upbeat: “The company is probably better positioned than it has been in years,” he states, “working with Fujitsu, should put them back in the race again.” Though as Felburn points out, “The process technology is an important factor, but whether Lattice has the architectural abilities or the sales channels needed to exploit this is still to be seen.”
Both Actel and Lattice know full well that it is not going to be an easy thing to reverse their fortunes. “Xilinx and Altera are obviously the guys to beat,” says Tsui but, like East, he argues that the company does not need to take them on in every area. As he sees it, “Programmable logic is no longer a homogenous entity, it has developed into something far more complex.”
With this sector now reaching some sort of maturity he, like many others, believes that there will be more niches coming to the fore, and aims to capitalise on these. “It's no longer a ‘one size fits all scenario’, customers want to pick a device that can better meet their requirements,” Tsui states. “With the market becoming so segmented we do not always have to look at the mainstream. We can choose areas that offer more explosive growth, where the big guys are not already locked in. Low cost FPGAs, for example, currently only account for 10 per cent of the market, but over the next three years it will account for over 30 per cent,” he predicts.
Just as Actel is, Lattice is employing flash technology in its newer generations of devices. “The non-volatile market has been incredibly specialised until now, accounting for maybe 6-7 per cent of the business generated, but with the advent of flash we are now able to solve the problem of re-programmability, so this should be at least 15-16 per cent of the TAM by 2008.”
Like the others, he sees a growing interest in programmable logic away from the mainstay application areas in communications. “In the past people would not even consider an FPGA for a consumer application,” he states, “now with the reductions in size and power consumption that these devices have seen it is becoming more attractive, epically as Asics are increasingly pad limited.”
It's not just consumer that is attracting attention, each of these companies is also putting a lot of effort into trying to break into the automotive arena. “In the old days, car models were made with exactly the same features, painted the same colour and so on,” says Actel’s East. “Each ‘Model A’ that Ford rolled off its production lines was identical to the last. Now everybody wants something different, so the idea of making an Asic for each of these options just does not make sense commercially speaking, programmable logic is a much more suitable candidate.”
The important differentiator that his firm sees here, is that coming from the military/aerospace arena, it is already used to the stringent regulations expected by OEMs. In East’s opinion the more rugged devices that his firm can supply mean that it will be able to tackle ‘under the hood’ applications as well as the infotainment and telematics tasks that its competitors are targeting.
Industry guru Jim Tully is another who believes that there is still hope for the industry minnows. He sees opportunities arising that could offer a way for them to stop the gap from widening any further. “There could be openings for the utilisation of embedded FPGA core technology in standard cell devices, thus offering an element of programmability to these chips. Because the area of silicon needed for such cores is relatively small, the inherent cost premium that FPGAs have over Asics won’t be so apparent. If they can partner with one or two large semiconductor manufacturers they may find another undeveloped niche which could generate some substantial revenue.”
“Xilinx has already done this with IBM, and I think this offers an interesting way for the smaller players to claw back some market share,” he concludes. A recent report by In-Stat has suggested that Embedded FPGAs could be an $800m business by the time 2008 draws to a close.
Lewis concurs with Tully on this: “Lattice tried to do this before, the problem was at larger architectures it would take up too much of the chip to make it viable. However, at 90nm or below it could start to make sense.”
Whether these companies can rise to the occasion or come up short is still uncertain. Luckily the sector is probably not yet at a stage of complete adulthood, and if all the corporate sales pitches are to be taken as gospel then there will still be a plenty of growth here.
iSuppli forecasts that the sector will by worth a little shy of $5bn by 2009, so just maybe a firm that manages to mine the right seams can still make the big time. One day perhaps the underlings might turn the tables on the established nobility, but it is going to take a lot of hard work and some crafty moves to get there. Nevertheless, here’s to the revolution.
Mike Green is managing editor of EPN