Naveed Sherwani, outspoken CEO of Open Silicon, feels it is time for change. In his opinion, "The development cost, time-to-market, and size of the chips don't just have a small effect on the total revenue that you can take for a specific project. Each causes an exponential drop in the return on investment, and dramatically reduces the value of an ASIC solution, until in the end it's not worth it." As he sees it, this sector needs to clean up its act, becoming more responsive, cost-effective, and operationally efficient. His firm, which effectively allows OEMs to subcontract their ASIC development, is trying to lead a revolution in the way that these increasingly complex and costly chip projects are brought to market. "The ASIC sector is not likely to survive if it carries on down the route it is going, because of the cost metrics involved," he states. "A traditional vendor can only claim 70 to 80% first-pass reliability, and schedule predictability is even worse (just 15 to 20%). Imagine if an airline or freight-shipping firm could only manage to be on schedule 15% of the time, it would go out of business." This is where Sherwani sees the problem. He holds the view that the ASIC sector has been underachieving for some time. "Any industry needs to have 90% predictability levels about when it will be able to get its job done. If it can't guarantee this, then something is clearly wrong," he thinks. "Look at the motor-car industry in the early years: hundreds of little manufacturers with no standardisation. Then there was a paradigm shift, and companies like Ford looked to improve the efficiency of production, thereby reducing cost, and the market expanded a hundred-fold," he comments. "Currently a lot of designs are never realised just simply because of the restrictions set by costly, inefficient business practices. By amending these faults, new application areas could suddenly be made viable."

Another issue that seems to really annoy him is the lack of communication that ASIC vendors have with their client base. "They need to be far more visible, letting customers know exactly what is happening on their projects: if they are on schedule, or if problems have arisen." In order to address this, Open Silicon has introduced an on-line schedule tracker, which allows customers to look on the Web to see how things are progressing.

Walden Rhines, CEO of Mentor Graphics, agrees with some of the points raised by Sherwani. At a recent event in California, he stated that, "the ASIC market is in the process of curing itself. The emergence of an outsourcing design model here is changing the whole business structure that it follows. These guys can utilise their expertise in chip development and leverage good deals with Far Eastern fabs thanks to greater economies of scale."

Rich Busch, director of ASIC products at IBM, certainly doesn't look to discredit any of these comments or deny a change in the business landscape, but at the same time he doesn't think that the effect of these trends should be over-exaggerated. Over the course of the last decade, the Big Blue one has held the number 1 position in this market for most of the time, and through its fab partnership with Chartered and Samsung, as well as its cross-discipline strength, it has won lucrative ASIC contracts such as the Sony PlayStation 3. "Actually, from IBM's perspective the signs have continued to be favourable. The number of designs started during 2006 was up by over 30% compared to 2005, and sales revenue for our ASIC products has witnessed double-digit growth." What he sees is that though the number of design starts in the industry as a whole is declining slightly, this doesn't give a true representation of the vigour still in the market. He states : "The value placed upon each project is actually increasing. The greater complexity and the higher level of software-support needed means that every design requires a lot more to bring it through to completion. Also, in many areas OEMs are making use of one ASIC in multiple versions of various products by taking a platform approach that reuses their IP rather than just starting a design from scratch each time. So the unhealthy state of the ASIC business is often greatly exaggerated."

"Market analysts will point out that the cost of a design is doubling with each new process node and the risk of respins increasing dramatically, but, at the same time, the need for our customers to differentiate themselves from their competition is greater than ever, so the value of ASIC solutions is still very apparent," notes Busch. He adds, "There are factors that are creating a lot of consolidation. Several companies have thrown in the towel when it comes to process research, refocused themselves on standard products, and abdicated their positions in the ASIC sector, but this has meant that our market share has continued to rise." He feels little room in this market for "also rans" now. "To be able to stay here you need to be able to combine manufacturing strength, fundamental research, and depth of intellectual property. The stakes have continued to be raised, and this is now clearly a "big-company" arena, where SMEs just won't be able to compete." All this goes some way to explaining what has occurred in the ranks of the ASIC industry's middle order. Some players, such as LSI Logic, are looking to withdraw from this sector and pursue other more fruitful markets, while others choose to concentrate on a specific niche, such as mixed signal or FPGA conversion.

With regard to vendors like Open Silicon, Busch sees their emergence as a "natural evolution". As he points out, "With the widespread availability of foundry services that we now see, it makes sense for third parties to offer ASIC solutions that can utilise this. They serve a purpose, but it is not where we are at. The dimensions they can bring to a design are much more limited. They can meet the needs of some companies, but the larger firms with challenging designs will not consider them."

He feels that though people have tried to downplay the future of ASICs, this sector is in better shape now than it ever has been. "In the past, customers would just make use of the current leading process node and follow it blindly; now they can look at a whole series of different options and work out what is the best in terms of NRE, unit cost, power consumption, etc., and match it to their specific requirements. Also, they can shift to another node when their technology requirements or business model changes. The choice is far greater now, and that means that companies can get exactly what they want."

There seems to be widespread agreement that changes in the way things are done are necessary, and though firms like IBM may not wish to acknowledge it, the charms of programmable logic are now becoming more profound at the lower end of the ASIC TAM. "We are going through a transitional phase, where new business models need to be considered," according to Gary Meyers, CEO of Synplicity. "In many cases, FPGAs now have a two-node advantage over standard-cell designs, as ASICs solutions have tended to stay at 130nm since the costs of going lower are too great." The advent of the Xilinx Virtex 5 and Altera Stratix III 65nm solutions has meant that far greater levels of performance, power efficiency, and silicon utilisation can be realised, and so the disadvantages that these devices had in the past are far less acute now. He sees other changes in the FPGA sector afoot, adding "There are new vendors entering the fray as the market gets larger and many of the fundamental patents here start to expire."

One new player in this formerly exclusive sector is Achronix. This firm manufactures ultra-fast FPGAs, operating at close to 2GHz, which it expects will allow it to capture a segment of the ASIC market that would have been inaccessible to programmable-logic devices until now. Another is Oregon-based MathStar, who's FPOAs (field-programmable object arrays) execute logic functions at speeds of up to 1GHz, bringing the processing benchmarks normally associated with ASICs into the programmable-logic arena.

It seems likely that in the future the ASIC arena is going to become a rich kaleidoscope of different commercial methodologies, with traditional vendors offering more service-led solutions to the big OEMs, outsourcing firms supporting the smaller players one smaller volume, less complex project, and programmable-logic firms making inroads into areas where development costs and time-to-market create pressures. Though the rules of engagement are different from those of the past, there seems to be no reason why firms that learn how to adapt quickly can't gain from this.