Handel Jones, founder and CEO of International Business Strategies (IBS), spoke at SEMIâ€™s Industry Strategy Symposium last week, focusing on key trends, factors impacting the growth of the industry and the migration to smaller feature dimensions. He is bullish about 2014 and industry innovation, but cautious about how quickly the industry will move to new technology nodes due to higher costs, and long design cycles. Overall, he said he believed semiconductor market growth this year will be slightly better than 2013, due in part to the strength of the global GDP.
Perhaps most surprisingly, he had a fair amount of uncertainly about 20nm. Â â€śWill 20nm be a high tech technology node and when will that occur?â€ť he said. â€śWeâ€™re tracking design starts and design completions and we see a few 20nm designs but not a lot. Frankly, whether 20nm will be big or not will really depend on two customers: one is Qualcomm and the other is Apple.â€ť Handel said â€śthere is a significant challenge in getting lower cost at 20nmâ€ť compared to 28nm due to a lack of increase in the gate density and the potential yield impact. â€śWe think 20nm, if it does go into volume production, it will not be in 2014. Potentially 2015 and maybe 2016,â€ť he said.
Similarly, Handel believes there will be a postponement of 16/14nm. â€śWe expect initial production in late 2016, beginning of 2017. Thatâ€™s for the SoC business. The FPGA markets will be different,â€ť he said. â€śThere will also be delays in 10nm. Delays mean you canâ€™t really go on the 2 year cycle or even the 3 year. I know people will vehemently disagree with that, but if you look at whatâ€™s really happening from a design start point of view and also the end customers, I think youâ€™ll agree with our conclusion,â€ť he said.
â€śIf you look at the reality of the industry, 28nm high-k metal gate went into high volume production toward the end of 2013,â€ť said, adding that they define high volume as 10% of the output. â€śIt took almost 4 years for 28nm high-k metal gate to go into high volume production. Now weâ€™re basically starting 20nm. Even if the fabs are ready what you have is the design cycle time. Preparing libraries and IP can take six months at least. Doing a complex design in 20nm can take you at least a year. Validating the design can take you another half a year. If itâ€™s a modem, and you need approval from the carriers, thatâ€™s another half a year. Even if the fab is ready, you start these things and itâ€™s two years,â€ť he said. â€śWe have an industry that is trying to adopt three technologies in three years. Itâ€™s impossible,â€ť he said. â€śItâ€™s not realistic from an infrastructure point of view, even if it the fabs are there, for three technology nodes to ramp in three years.â€ť
Handel said that application processor (AP)/modem design can cost about $450-500 million in 16/14nm, with a timeframe of around 18 months. â€śYou need 10X revenue so for that design, so if youâ€™re spending $450 million, you need $4.5 billion in revenue. A few companies can get that, but not many,â€ť he added.
â€śThe economics of the industry are forcing changes. Youâ€™ve seen them already. The long ramp up time for 28nm HKMG, and 20nm with double patterning is clearly a major challenge from a technology point of view, and a bigger challenge from a cost point of view. FinFETs will be an even bigger challenge. Intel is having delays in their 14nm FinFETs, whether in high volume at 22nm, how will companies that have never done FinFETS before, how will design companies that have never designed in FinFETs before, how will they ramp faster?â€ť he asked.
Â Not surprisingly, Handel also had a dim outlook for 10nm. He estimates that 10,000 wafers/month at 10nm will cost more than $2billion. â€śIf you want to install 40,000 wafers/month, itâ€™s going to be an $8 billion bill. If you want to install 100,000 wafers/month, itâ€™s going to be $20 billion. Even before you get to 450mm, itâ€™s going to be significantly more capital intensive,â€ť he said.
Just looking at the location of the headquarters of semiconductor companies, he said the U.S. was still strong, but there was also strong growth from Korea â€“ mostly in the form of Samsung â€“ but also China and Taiwan. â€śWe see a relatively flat Europe and then a continuing decline in Japan. In fact, we donâ€™t see Japan strengthening unless we see some major changes,â€ť he said.
That also has an impact in terms of the technology requirements. In terms of minimal dimensions, Handel most of the advanced technology designs are in the U.S., with advanced technology defined as being 28, 20 and now starting 16/14nm. â€śIn developing countries, many of the designs are still at 40nm. 28 is a new technology and the next technology after 28 is going from polysilicon up to high-k metal gate,â€ť he said.
Handel also sees uncertainly in the use of FinFET devices due to higher wafer cost. â€śWe see quite a few new designs. The problem again is the cost per wafer. For 28nm, we have about $2600 and for 20nm we have about $3200 and for 16/14 we have about $4000. You now have this increasing cost per wafer and can you get the higher gate density and can you also get higher parametric yields?â€ť he asked.
Handel said the gate utilization is an issue because of limitations of the design tools and parasitic effects. â€śThe other factor is parametric yields, which are strictly tied into leakage control for the 20nm and of course for the 16nm FinFETs,â€ť he said. â€śYou can break this. Intel has shown that it can be broken and of course thatâ€™s an excellent achievement. But, itâ€™s based on very high design costs, potentially $1 billion per design, so you need $10 billion in revenue. It also takes a number of years,â€ť he said. He noted that, in the smartphone market, designs move very fast. â€śYou canâ€™t make that kind of investments in designs.â€ť