Malcolm Penn, market analyst, founder and ceo of Future Horizons

Semiconductors are one of the largest industries in the world, generating revenues in excess of $200billion a year, yet the industry almost always lacks confidence in its future. You wonder why when the industry, over the 40 years or so of its existence, has grown consistently by 15% per year. Why does Malcolm Penn think the industry is so pessimistic when it has such a track record? "It often doesn't matter who you talk to, they say business is good today, but tomorrow will be bad. Semiconductor revenues have grown by about 0.8% per year since 2000, which sounds bad. That's the driving factor behind the pessimism. For an industry that used to grow at 14% a year, that's close to the end of the world as we know it." But Penn says things aren't as bad as they may seem. "Mathematically, the figures are correct, but you have to look at why they come out like that." One of the problems is the start point was at the peak of a semiconductor cycle, while the end point was in a trough. "We couldn't have a worse situation," he noted. "If you moved things plus or minus a year, growth would come out at 8%; not fantastic, but better than 0.8%." Yet unit shipments grew by 10% in the same period. "That's the long term average," he continued. "If unit growth is 10%, then we have an average selling price (asp) problem. That puts things into a different perspective." Penn blames a combination of circumstances for the drop in semiconductor asps. "Conversion of many fabs to 300mm wafers is one factor," he said. "The industry went to 300mm to reduce manufacturing costs, but manufacturers appear to have given away those savings. In fact, the asp for some devices declined by 40% after they moved to 300mm." He also pointed to the 'carnage' in the memory industry. "This sector was one of the first to go to 300mm and ended up making 2.5 times as much silicon. The problem was the market couldn't absorb that, so manufacturers had to get the price down or dump products." It was a chain effect, Penn said. "When one manufacturer moved to a lower cost structure, the rest had to follow whether they needed the capacity or not. It has taken four or five years to sort out the problem and now memory prices have doubled." All in all, said Penn, the industry lost the asp advantage that would have been expected with a move to the next technology node. But he believes the asp problem is over. "We're back on the curve; the asp problem of the 2000s couldn't go on forever." Another reason for Penn's optimism is the state of the global economy. "The world's GDP may have shrunk by a couple of percent, but it's still at the level it was two years ago. Recession hasn't taken demand away, just postponed it." And he believes demand will return shortly, with enterprises fuelling growth, rather than consumers. "There are a lot of old pcs in use out there which need to be replaced. This will put huge demand on workstation and server suppliers. It's already started, with demand for pcs growing more than anticipated." When it comes to manufacturing, Penn believes the move to 450mm wafers will happen, but sees growing problems with the pursuit of Moore's Law. "450mm will happen; no question. It's working on pilot lines and Intel, TSMC and Samsung want it to happen. But physical problems with transistors will affect how quickly we follow Moore's Law. Today's problems at 45nm are the tip of the iceberg. Transistors are not behaving in the way they used to. Today, this 'misbehaviour' affects 10 to 15% of the transistors on a chip; at 16nm, it could be up to 50%. That will affect design methodologies and change our world significantly." Even after 50 years of semiconductor manufacture, Penn believes the industry has yet to get its act together. "If you build planes, for example, you will get orders from airlines and commitments for the future. With the chip industry, there's no commitment at all. While the level of investment is huge, no company has any long term orders. The industry must learn to plan capacity better and to make it more of a partnership, rather than confrontation. There have to be changes in the model. It's been done at the process R&D level; now it needs to happen at the equipment level." One reason why he advocates this is that he sees potential problems with the supply of advanced semiconductors. "There's really only one foundry offering a 45nm process and that's a weak link." For this reason, he believes the 'fab lite' model, currently the manufacturing approach of choice amongst IDMs, is flawed. "We need a manufacturing alliance similar to the technology alliances that have been created. What's wrong with the concept of a fab owned by a lot of little customers? The industry works together on process R&D, why not work together in manufacturing? I think there's a good chance this will happen and Crolles was a microcosm of this. If it works for foundries, why shouldn't it work as a joint model?" Plans are well advanced for Penn's International Electronics Forum, being held in Dresden from 5 to 7 May 2010. The event brings together industry leaders to discuss the key issues. This year's event is titled 'Thrive or Survive: Going for Gold in the post recession economy'. Amongst companies presenting are TSMC, GlobalFoundries and Freescale. For more on the event, go to http://www.futurehorizons.com Malcolm Penn, Founder and ceo of Future Horizons Penn has worked in the electronics and semiconductor industries for more than 45 years in Europe, the US, the former USSR, Japan and Korea. An early pioneer of pan European research and product development during his time with ITT?Europe, Penn's industrial experience has involved him in all aspects of the management, manufacturing, marketing and use of electronic components, particularly semiconductors.