Analysis: Fabless firms feel the pinch
In a tough climate, E&T finds that many fabless semiconductor businesses are drawing back from the latest and most expensive processes.
Faced with the double whammy of rising costs and falling prices, fabless chipmakers are looking for new directions to go so they can start making money again. At the GSA and IET International Semiconductor Forum, the heads of some of the European chipmakers told of the problems they face with increasing complexity and a punishing business environment.
Responding to a question about how business models have evolved in the silicon industry, Joep van Beurden, CEO of Cambridge-based radio specialist CSR, remarked that the model has hardly changed in some 40 years: "Fundamentally, we sell our silicon by the square metre. You jump to the next node and then drop the price. That makes technology accessible but I would argue that we are taking this a bit too far now.
"The semiconductor industry resembles the bathroom or kitchen tiles business. We all sell by the square metre. The difference is that the tiles business has better margins than where we sit. The question is how do we get out of this and capture more of the value that we provide with these innovations."
Against an environment of falling prices, the level of investment needed for a chipmaker that can deliver products based on the most advanced nodes has spiralled upwards. Stan Boland, president and CEO of Icera Semiconductor, which makes chips for cellular communications, said the company has now raised $240m, "which is a lot for a fabless start-up".
Boland argued that fabless start-ups are faced with a choice. Focus on relatively small niches where older, cheaper processes can work, or make a play for a much bigger market and raise much more funding. "Put enough wood behind the arrow and success will come. Through sheer force of financing, the reward will come," he declared.
Price pressure is reducing the number of players who can go after large markets such as cellular, Boland acknowledged. Customers will often negotiate based on their understanding of suppliers' costs, which has squeezed gross margins from the historical average of 60 per cent. "We could be in a 40 per cent gross-margin business now."
The problem, Boland explains, is that it forces the smaller, often nimbler players out of the running. "Rather than a coral reef of innovation it creates more of a deep-sea environment where the coral reef has gone."
That environment, Boland said, is one patrolled by large, hungry sharks. Smaller fish do not last long. "The question is how do we create valuable companies in that deep-sea environment?"
Tudor Brown, president of ARM, insisted that there are applications that require what the latest semiconductor processes offer, which means high levels of investment. "It's the only way to get very high levels of integration. But it doesn't mean every product has to chase the nodes. There is a bifurcation going on. It's something that people are surprisingly reluctant to talk about but it is healthy to look at those two tracks."
Simon Atkinson, CEO and founder of mobile-TV chipmaker Mirics Semiconductor, said his company has taken the view that chasing advanced processes is not going to work: "As an industry, we have more raw silicon technology than we know what to do with now. Mirics is using older process technology to get the advantages of greater capital efficiency rather than just driving down the process nodes to shrink the die."
David Baillie, CEO of Cambridge Semiconductor, which makes power semiconductors, agreed: "Our view is that we avoid the bleeding edge."
Baillie explained that his company focuses on 0.35µm technology, more than ten times less dense than the upcoming 32nm processes. Although capacity for these older nodes has shrunk since they peaked in the late 1990s, Baillie argued that the lower capital cost has led to renewed interest from foundries: "Foundries are actively seeking out spaces that don't require the bleeding edge.
"We are in a market where, previously, the most complex component they had in there was a bipolar transistor," said Baillie, adding that opportunities for simpler silicon are turning up all the time.
"Looking around the room, I don't see a single intelligent lightbulb. Those standard lightbulbs are just 5 per cent efficient. It's insane and immoral to be using lightbulbs that are 5 per cent efficient. That will change over the next ten years. And they will have significant semiconductor content in them, compared with zero today."
ARM's Brown agreed: "There is a lot of unsexy stuff ahead of us. But there is a lot of business in front of us there too."