SAN JOSE, Calif. — After a tough 2013, John Daane, chief executive of the FPGA vendor Altera, shared his views on the outlook for semiconductors and engineering in a wide-ranging interview with EE Times.
I started off asking about his opinions on the slowing and eventual end of Moore's Law in the next 10-15 years at the 7-5 nm node. The near term looks upbeat, because scaling will continue "for several generations," offsetting rising fab costs, Daane said. "We need to ride this curve and for the next ten years we can do that."
In the longer term, "we have to switch to a new switch, [because] you can't split the atom, so we have to go to a new material other than CMOS." He recalled similar fears when he started his career 17 years ago. "Everyone said one micron was the end, but we found new materials, [so] fundamentally, I believe Moore's Law continues."
Some say the move to 3D chip stacks will give a boost to high-end chips about the time Moore's Law starts slowing. Daane expressed some skepticism about chips stacks, at least in the short term, even though (or perhaps because) Xilinx has shown leadership in so-called 2.5D stacks that link die places side by side on a silicon interposer.
"The cost is still too high for the silicon substrates, so you will see it in in niche products for much higher performance or functionality. Those will be first adoptions. When it goes into handsets is when it hits its stride," he said. "We've been doing a lot of research [in the area], and you will see products from us." He would not say whether any of them will ship in 2014.
The good news: Altera had one of its lowest turnover rates in its engineering ranks this year after a similar low rate in 2012. This year, the FPGA vendor hired several hundred college grads and experienced engineers, mostly in the US. "Unemployment is down in Silicon Valley but still higher than it was in 2010." However, "we're not going to be graduating enough engineers in the US."
The Semiconductor Industry Association funds some research in EDA, materials science, and other chip-specific areas to bolster grad programs. Still, "US masters and PhD candidates in STEM fields are more than half [that of] foreign nationals, so we're not getting US nationals interested, and that [indicates issues] much earlier in schooling."