Silicon has enjoyed some serious staying power. For going on 60 years, it’s been the semiconductor of choice at the heart of transistors—the tiny switches that power the Information Age. A valley has been named after it. Many billion-dollar empires have been forged from it. And now—look away, silicon—it may finally be on the verge of being replaced.
The cool new semiconductor on the scene is gallium nitride, or GaN. Transistors made out of GaN can turn on and off faster and withstand higher voltages than those made from silicon. When you add these and a handful of other qualities together, GaN transistors—each slightly bigger than a grain of salt and sitting on tiny circuit boards as pictured above—should allow companies to make smaller, faster, smarter, and more power-efficient products. While GaN transistors are not going into PC chips today, they are being used to improve performance in prototypes of self-driving cars and virtual-reality helmets and are paving the way for entirely new consumer-electronics and medical devices. Think wireless charging devices and X-ray pills. “Here’s a place where we actually have a new material and are looking at some applications that have not quite existed before,” says Stephan Ohr, the director for semiconductor research at Gartner.
The most vocal champion of the GaN revolution is Alex Lidow, a physicist and chief executive officer of Efficient Power Conversion (EPC) in El Segundo, Calif. Lidow, 60, has spent much of his career seeking a material that can perform operations faster and survive harsher conditions than silicon. Over the past two decades, he and dozens of other scientists have toiled to take GaN transistors from the experimental stage to the point where they can be produced at standard semiconductor manufacturing plants. EPC’s transistors are in only a handful of products. But over the past 18 months, production methods have improved enough for the technology to go more mainstream.
“For the first time in 60 years, I’m able to say, ‘I can make this thing better, and I can make it cheaper,’ ” says Lidow.
In 1977, Lidow co-invented a type of silicon-based transistor known as a power MOSFET (metal–oxide–semiconductor field-effect transistor). This device, which costs about 17¢ a pop today, has been used to switch electronic signals in a huge range of products, from server power supplies to washing machines. One of the world’s leading suppliers of MOSFETs is International Rectifier (IR), a company founded by Lidow’s grandfather and father that became a $3 billion business on the back of his invention. Much of the excitement around GaN transistors centers on their ability to replace MOSFETs and revolutionize the field of power management.
Because GaN transistors can be turned on and off faster than silicon-based ones, it’s possible to control them more precisely. As a result, companies can make more efficient, smaller power conversion systems for use in industrial settings and the home. “Power supplies and adapters are all dying for a new, more efficient technology,” says Umesh Mishra, a professor of electrical and computer engineering at the University of California at Santa Barbara. A wholesale move to GaN-based transistors in these types of products, Mishra estimates, would save about 40 nuclear power plants’ worth of electricity worldwide per year. “When talking about climate change, people often say you can’t tackle waste without spending more money, but here is a case where that is not true,” he says.
The U.S. Department of Defense and the Department of Energy have sponsored Mishra’s work on GaN transistors for two decades, and they’re already used by the military in radar and communication systems. In 2007, Mishra co-founded Transphorm, which sells GaN transistors for power conversion. Customers include India’s Tata Power Solar and Japan’s Yaskawa Electric. Transphorm is backed by more than $125 million in funding from Google Ventures and Kleiner Perkins Caufield & Byers. Semiconductor heavyweights such as Infineon Technologies and Panasonic are also producing GaN transistors for power applications.
Lidow, unlike these rivals, is chasing novel, higher-profit uses for GaN transistors. “They’re going after a big existing market,” he says. “We’re creating new markets and actually generating new business.”
In 2007, Lidow resigned as CEO of IR. He took the fall, he says, when $14 million worth of dodgy accounting turned up in the company’s Japanese business. He couldn’t get another job. “The recruiters told me I was toxic,” he says. So he decided to start his own company, EPC, and hired researchers who had been developing gallium nitride under his watch at IR.
Lidow and Archie Hwang, a Taiwanese businessman and owner of a semiconductor plant, each put about $25 million into EPC and together own almost 99 percent of the company. IR and Lidow sued each other and then settled in June 2013. EPC agreed to pay royalties on the GaN transistors it sells. (Infineon acquired IR last year. IR didn’t respond to requests for comment.)
This year, EPC expects to sell millions of its GaN transistors, Lidow says, priced from 40¢ apiece to a few dollars. They’ll go into self-driving cars, where the high on/off switching speeds of the transistors help sense obstacles about 10 times faster than silicon transistors. Telecommunications equipment makers already use EPC’s transistors to boost the signals of their wireless base stations. Consumer-electronics makers have put EPC’s GaN transistors into their virtual-reality helmet prototypes. One company has built a GaN-based pill, expected to go on sale this year, which takes X-rays of people’s colons and saves them from going through colonoscopies. Scientists were able to shrink the X-ray machinery to fit into the pill because the GaN transistors are so efficient. Lidow says he can’t reveal all his customers for now. Among those he names are Microsemi, a chip and electronics maker in Orange County, Calif., which is using GaN transistors in satellites, and WiTricity, a startup out of MIT, which has developed wireless recharging systems that can work at distances of several feet. Another customer is FINsix, which has built a laptop charger the size of a cigarette lighter.
Gartner’s Ohr emphasizes that overcoming the inertia of the semiconductor industry won’t be easy. While GaN has attractive properties, its advocates need to prove it has both dramatic performance advantages and major cost benefits. When it comes to things like PC and smartphone chips, industry inertia means silicon transistors are unlikely to be replaced for a very long time. But Lidow and others still see plenty of opportunity. Over the next few years analysts estimate that GaN transistors will take over about $1 billion of the silicon-based transistor market, and Lidow thinks there’s $300 billion up for grabs from there. “I have a vision,” he says.
The bottom line: Gallium nitride promises to replace silicon as the semiconductor of choice in transistors.