Silicon chips have had a decades-long run as the foundation for modern electronics. But a new kind of chip, based on the compound material gallium nitride (GaN), promises to unseat silicon because it has higher performance, less power consumption, and lower cost.
That might sound crazy, given the economies of scale for silicon, which has been an almost perfect semiconductor device that can both conduct or block electrical signals. Many people scoffed in early years at alternative materials such as gallium arsenide, which proved to be too expensive. Gallium nitride’s assault on silicon is just in time for the 50th anniversary of Moore’s Law, the prediction by Intel chairman emeritus Gordon Moore that chips would double the number of transistors every two years.
If the advocate for gallium nitride, Alex Lidow, is successful, it could be a huge and profitable disruption where consumers win in a big way. We could see cheaper and better electronics products, like wireless power, video game gesture control systems, and self-driving cars that can sense their surroundings.
Lidow has already changed the chip industry once before in the late 1970s with power MOSFET chips. Those chips resulted in a 15 percent decrease in the global cost of energy. For example, today’s energy efficient home appliances, air conditioners, energy-efficient lighting, and high fuel-economy automobiles are all built with the MOSFET technology that Lidow helped create.
Lidow is now running a five-year-old startup called Efficient Power Conversion, a Los Angeles company that is designing GaN chips. And he’s trying to change the semiconductor industry, one GaN chip at a time. Lidow believes that a new type of semiconductor material, GaN (Gallium Nitride) will soon dominate the $31 billion power management semiconductor market. He has the patents, the supply chain, the customers, and the moxie to do it. His ultimate goal is for GaN to overtake the broader $300 billion semiconductor market.
“This is not a shot at the bow,” Lidow said in an interview with VentureBeat. “This is a shot at the water line of silicon. Other challengers had an advantage in cost or performance, but not both. In my lifetime, there’s never been a case of better performance and better cost at the same time. If Steve Jobs had held up an iPhone in 2007 and said this is faster and cheaper than your old cell phone, there would have been even faster adoption of smartphones.”
He added, “We are offering a radical performance departure. It is 10 times better or more, and we are just at the beginning of Moore’s Law on that. We can build it at a lower cost. What are people waiting for? They are waiting for the risk to go away.”
He said the Lidow says that the advantage of gallium nitride is the ability to manufacture it in the same factories that produce silicon chips. In contrast to gallium arsenide, the GaN chips can be processed without expensive machinery or special processes. In fact, the GaNs chips can be made on top of a silicon wafer, without the expensive packaging for protection. The GaN is sealed up on top of the silicon, and the silicon itself is not electrically active.
Packaging for silicon chips can often be half the cost. That packaging remains necessary for silicon chips in order to protect the chips from contamination or short circuits, since the entire device is electrically active. Gallium nitride chips don’t need that same protection. In fact, about 15 years ago, researchers figured out how to grow gallium nitride crystals on top of silicon. By growing a small layer, it isn’t that expensive.
“You don’t need a package,” Lidow said. “Half the cost is gone, and it’s a third of the size of a silicon chip that does the same thing. So you can see how we can make it for a lower cost than silicon.”
At first, as it was working the kinks out, EPC focused on premium markets for its chips. Now it is going after major markets.
“Gallium nitride is a superior semiconductor,” Lidow said. “I don’t think anyone would dispute that. Electrons are much more efficient in GaN than they are in silicon. But like many materials, GaNs in the past has been much more expensive.”
It was relegated to power amplifiers in cell phones, radar systems, and high-end military systems. But Lidow and his cofounders Jianjun “Joe” Cao and Robert Beach say they figured out how to attack both performance and costs.
Lidow had a bit of a difficult start with GaNs. His former chip company, International Rectifier, fired him in 2007 after it found some accounting regularities in a Japanese subsidiary. (Lidow noted that the irregularities were limited to that one subsidiary, and the board thought it could run the company better than him). Then, after he started EPC and hired some International Rectifier employees, the company sued in 2009, alleging he stole employees and trade secrets. The suit was resolved a couple of years ago, and Lidow said, “I hired employees, but I didn’t steal trade secrets.”
After five years of tinkering and lining up manufacturing, Lidow said, “We are exploding on the marketplace this year. Next month, we will introduce a series of products that are higher performance and cheaper in price than silicon. This has not happened in the history of silicon.”
At EPC, Lidow believes that GaN chips can make practical Google’s driverless cars, cool virtual reality, better space exploration, colonoscopy capsules, and high-performance wireless power. That’s a bunch of possible breakthroughs.
But rather than just replace silicon. Lidow is targeting GaN chips at applications that didn’t exist five years ago, such as wireless charging of devices without cords, sensing systems for autonomous vehicles, and high-speed mobile communications.
All of the Google self-driving cars and other autonomous vehicles use EPC’s eGaN chips in their Light Illuminated Detection and Ranging (LiDAR) systems, or remote systems which sense the environment around them akin to radar. They illuminate an area with a laser and analyze the reflected light. Google uses LiDAR in the spinning rotors atop its self-driving cars.
With silicon, the chips are slower, so the accuracy is limited to a range of about ten feet, in terms of resolution. By contrast, with GaNs, the devices are accurate to within a couple of inches, Lidow said. You have to be able to detect the reflected light and do it very quickly.
“All of the Google cars use our eGaN chips,” Lidow said. “Car makers are looking to replace all of the sensors on the car with a couple of LiDAR systems.”
LiDAR is expected to become a $1 billion market by 2020, according to researcher Markets and Markets.
GaNs also work well inside wireless charging systems, which are just about to explode on the market and replace power cords. The wireless chargers on the market use either tightly coupled technology, where you have to closely align or plug in a device to its charger, or a loosely coupled one, where the device just has to be near the charger, maybe 15 centimeters or less. The latter is preferable, and Lidow said it is best done with eGaN chips.
Lidow believes that car companies will introduce power-charging systems in cars, with a loosely-coupled connection between a phone and a charging pad. Overall, wireless charging systems are expected to be a $1.7 billion market in 2015, according to market researcher IHS.
“Eventually, we’ll see power cords go away in the house,” Lidow said. “We have range anxiety now with our devices. We need to be untethered completely, and that’s coming.”
Lidow also believes that eGaN chips can be used to provide more efficient power to communications chips, making much more powerful 4G LTE systems. That’s going to lead to some high volume shipments of chips by 2016, Lidow said.
“That’s a billion-dollar market for our products that is developing very quickly,” Lidow said. “We’ve got a few markets that didn’t exist before GaNs power devices. But we can also replace existing silicon chips.”
Lidow isn’t free to talk about customers in other markets. But based on what Valve and HTC have said about their Vive virtual reality system, it looks like it will use GaN-based LIDAR systems to sense the environment in a room and build a map of where a user with virtual reality goggles could move without tripping over the furniture. That LIDAR could be an interesting part of a sensing system that enables you to interact with games using your body or hand gestures. Right now, Microsoft uses an expensive arrangement of cameras to accomplish the same thing with its Kinect gesture control system in the Xbox consoles.
Augmented reality goggles (think Microsoft’s HoloLens, in my own opinion) could also use LIDAR.
“The real experience of 3D virtual reality is to interact with your environment, not have a screen wrap around your eyeballs,” Lidow said. “The awareness of your surroundings is an integral part of the experience. You want to interact with your environment. You can do extraordinary things like remote surgery. It relies on gallium nitride, because of its intrinsic speed compared to silicon.”
Lidow faces a lot of skepticism that he will have to overcome through execution.
Stephan Ohr, an analyst at Gartner, said that Lidow showed him a module using MOSFET power transistors that can be displaced by GaN transistors and increase the electronics density of the module by 60 percent. Ohr said in an email that he “absolutely” believes that GaN will find a place in high-frequency applications and displace silicon in certain applications. But will it render silicon obsolete?
“I’m not so sure about that,” Ohr said. “Alex is a wonderful proselytizer, to be sure, and his enthusiasm is infectious. Alex was not particularly enthused with analysts like me who believed that GaN use will grow primarily by displacing existing silicon transistors in existing applications. Rather, he is excited by brand new applications — automotive Lidar or wireless charging (using A4WP) — where silicon transistors had barely begun to proliferate with GaNs.”
Ohr estimates that GaNs chips are a $45 million market this year. But it could grow to more than $200 million by 2018, assuming a 2 percent replacement rate for silicon by that year.
“This will likely shoot up dramatically (the “hockey stick” we look for) in 2019 and 2020, as penetration rates increase in power supplies and the newer applications,” Ohr said.
Silicon chip makers may say they aren’t threatened. But there are a dozen or so GaN startups moving into the market, with funding from some very big companies like Intel, Samsung, and Panasonic. Texas Instruments has a GaN solution that uses EPC’s chips. Panasonic also re-introduced the Technics brand based on EPC’s chips, based on a brochure that Panasonic showed off in January.
“We’re the cornerstone of that introduction,” Lidow said.
Back in January, mobile communications firm Infineon Technologies purchased International Rectifier, the original power chip company that Lidow’s father, Eric Lidow, founded. Alex Lidow believes that GaNs are most likely to disrupt the business of Infineon. At International Rectifier, in the 1970s, Lidow introduced power MOSFETs, which contributed to a net 15 percent reduction in global energy costs. International Rectifier introduced its power MOSFETs in 1979. By 1983, it had 43 competitors. Today, power MOSFETs are a $16 billion market, and Lidow wants to disrupt it.
Lidow will have to stay ahead of some powerful companies. Intel, the king of silicon and the world’s largest chip maker, is paying attention. It isn’t yet making gallium nitride chips, and it declined to comment on EPC or the displacement of silicon.
But in a research paper published in June at the VLSI Symposium, Intel researchers said, “These results make GaN MOS-HEMTs attractive for realizing energy-efficient, compact voltage regulators and radio frequency power amplifiers for mobile system-on-chips(SoCs) . This work shows, for the first time, that the application space of GaN electronics can be expanded beyond the existing high-voltage power and RF electronics (e.g. automobile, power conversion, basestation, radar) to include low-power mobile SoCs.”
But he can take heart in electronics history. The chip industry has been disrupted before. IBM’s Bernie Myerson introduced silicon germanium to replace silicon in high-performance communications chips. Silicon germanium, a hybrid of silicon and germanium, is now fairly ubiquitous because it brought about a 20 percent improvement in power efficiency. GaN’s practical improvement is 10 times, and its theoretical improvement is 1,000 times, Lidow said.
Lidow is using a similar strategy to what Myerson used. EPC creates sealed gallium nitride wafers in its own facility, using a metal organic chemical vapor deposition chamber. These are the same machines used to make blue light-emitting diodes, and there are several manufacturers of these machines. Then EPC passes its treated wafers to the manufacturing line of Episil, a contract chip manufacturer in Taiwan, which completes the processing in a standard silicon chip production factory. That keeps the cost of switching to GaNs low, Lidow said.
One way to get ahead of rivals is to get as many designs into the market as possible, and to write textbooks for chip designers on how to use GaNs. EPC is pursuing both paths.
“We want to make sure the textbooks are based on our specifications and our performance,” Lidow said.