In the 70s, Jerry Sanders, chairman of semiconductor manufacturer AMD, had uttered the famous politically incorrect statement, “Real men need fabs.” Those were the glory days of both Clint Eastwood and the chip-making industry, where any company in the Wild West of chip fabrication worth its silicon needed to own its captive processor plant, or so the thinking went.
Today, should that thinking still apply to India, planning to enter the commercial world of fab (industry term for a fabrication plant making silicon chips for computers, cell phones and other electronics)?
“Building and running a fab is a complex business that is very sensitive to utilisation and improvements in technology,” says Satya Gupta, chairman of the Indian Semiconductor Association. “Somebody who knows the fab business has to run it, not the government,” he adds. The Ministry of Communications and Information Technology seems to agree and is banking on private companies, which have been invited to bid for a project by enticing them with subsidies worth up to 25 per cent on capital expenditures. Among these companies are GlobalFoundries, STMicroelectronics, Infineon Technologies, Russia’s Sitronics, JSC and a consortium comprising Jaypee Associates, IBM and Tower Semiconductor. The ministry says a decision is imminent.
Still, experts say starting a fab is not for the faint-hearted. This is because the semiconductor industry lives and dies by Moore’s law (postulated by Intel’s legendary co-founder and chief executive, Gordon Moore), who, in the 60s, predicted the number of components you could put on a chip would roughly double every eighteen months due to rapidly evolving technology. This fact is the reason we’ve gone from phones the size of bricks to mobiles that stream movies.
Powering these phones are chips that today cost between $5 and $100 and measure between 90 and 28 nanometres (a human hair is 75,000 nanometres in diameter). So, it is difficult to imagine how one can put anything on something this small. But it happens, through a chemical, photographic method through where the processor is imprinted onto silicon. The trick for a fab today, including anything started in India, is to try and decrease the size of the chips and maximise the diameter of the wafer that houses these.
Doing so means India needs to be on the cutting edge of fabrication technology — like those belonging to Taiwan’s TMIC, the best of the breed, one of which costs $7 or $8 billion to build. Constant equipment upgrades for increasing wafer sizes mean being ready for higher capital expenditure — in 2005, research outfit VLSI estimated moving from 200-mm to 300-mm wafers a decade ago cost the industry $11.6 billion. And, running one isn’t easy. “If you operate at less than 80 per cent utilisation, you can lose $1 million a day,” says Gupta of SIA.
This is why China’s fab ascent in so astonishing. It has a $43.5-billion semiconductor industry and represents 47 per cent of the global consumption market. PricewaterhouseCoopers (PWC) estimates the industry in China would change from one with a share of less than two per cent of the global fab industry in 2003 to one accounting for a share of about 10.7 per cent by 2015. During the past year alone, the country increased its net number of fabs in production by 13 — a nine per cent increase in capacity, according to Raman Chitkara, an authority on the Chinese semiconductor industry at PWC. China saw 16 initial public offerings related to semiconductors in 2011. “The quality of Chinese fabs is absolutely world-class,” says Vikas Jain, co-founder of Micromax, whose phones use processors made in China.
But amidst this stellar growth story, China has an increasingly visible Achilles’ heel—something India should learn from. PWC’s latest report says for Chinese fabs, “revenue growth will lag significantly, as its foundries continue to compete on price rather than leading edge technology.” In other words, China is still churning out more commoditised chips than high-value ones, something India should not emulate. According to Dan Hutcheson of VLSI Research, most chipmakers are unprofitable and ultimately, consolidation would reduce their number down to a handful, perhaps just one, which may not bode well for China.
This is why today, companies are increasingly adopting a fabless’ model, one in which production by design firms like Qualcomm and NVIDIA, is farmed out to chip-making foundries’ such as TSMC, UMC and Samsung.
So, does this mean real men don’t need fabs’? Perhaps, but for India, not having a fab compromises research, as new designs regularly need to be shipped abroad to be tested, leading to crippling delays.
The good news for India is Moore’s law may finally reach its logical conclusion in the next decade. “The actual manufacturing of a chip, because everything is so close together, is very difficult,” says VLSI’s Hutcheson, in an interview with SiliconValleyWatcher. Design has evolved from being a “materials game” to a “transistor design one” he says, since a key objective these days is consuming less power. That can only be achieved by physically spacing transistors more efficiently.
This is something India has always excelled in, given its advanced and thriving integrated circuit design and embedded software industries.
Read the next part of the series tomorrow on chip design and how it, along with aspects of the recent electronic policy, can be a game changer for India