Get email delivery of the Cadence blog featured here
This is the story of how three rival Ph.D. engineers joined forces to start a company to attempt to disrupt power-management design in mobile devices.
The moral of the story (spoiler alert) is, you can create a semiconductor startup in Silicon Valley these days with no experience; it just doesn’t look or feel like the ones in the TV show “Silicon Valley.”
Harvard graduate and co-founder Wonyoung Kim came to Cadence May 20 to discuss Lion Semiconductor, the power management IC (PMIC) company he and some fellow Ph.D.s created two years ago. In a self-effacing but inspiring 45-minute talk, he described how the team went from concept to tapeout, found investors, and learned enough fast enough to fill a business case study.
“Luckily it’s still very possible to start a semiconductor company,” he said.
Design Tradeoffs Today
The story actually starts on the PC board. There, power-management capability is vital but physically demanding, taking up a lot of space, especially in a relative sense on small boards for mobile systems.
Most power-management ICs—which take in a voltage from the source and generate multiple voltages to power different chips within the system—require discrete inductors and capacitors. Some, like low-dropout regulators (LDOs), don’t require PCB inductors but have low current efficiencies (topping out at around 33% for a 3V-to-1V converter), Kim said. The alternative—a switching regulator—has efficiencies of 80-90% but requires PCB inductors and, therefore, is bulky and expensive.
Kim and co-founder Hanh-Phuc Le, who was getting his Ph.D. from Berkeley, had competing papers about how best to design better power-management ICs. They met at International Solid State Circuits Conference (ISSCC) a few years ago and decided to join forces to create, with another Berkeley Ph.D., John Crossley, Lion Semiconductor.
Lion’s solution “brings the best of both worlds,” Kim said. “It looks like an LDO from the outside because you don’t need a PCB inductor, but has comparable efficiencies to a switching regulator.”
He and his colleagues won’t share their secret sauce of course, but have a “ton” of patents and feel they’re in a reasonable position to build a business, even though as of yet they have no paying customers.
“We thought PMICs should be designed a certain way. We could persuade large companies to do it, but realized not many large companies are going to listen to fresh Ph.D.s,” he said.
He said Lion Semiconductor, in addition to the patents, has a different process, package, and circuit design.
“The best way to protect yourself is for the technology to be very different. It creates a big enough gap that it’s hard for a large company to decide to do the same thing until they see our success in the market, by which time we’ll have size and scale for better foundry deals,” he said.
Real Estate Opportunities
Kim claims Lion’s technology can help reduce power-management footprint on the board by up to 3X and board thickness by 1.5X.
“The thickest components on a mobile device in PMIC solutions are the inductors, not the silicon,” he said.
While acknowledging, “we are definitely not in a position to give advice,” Kim walked the audience through many of the lessons he and his colleagues learned—lessons you more likely would hear from a grizzled veteran of decades in the semiconductor startup trenches.
He noted that a semiconductor startup is more likely today in the analog/mixed-signal/power sectors because of the relatively lower overhead and general reliance on older, more cost-effective process technologies. A digital SoC or processor startup would require at least $100 million, 100 engineers, and a 10-year exit plan, he added.
That said, a startup still needs either enough differentiation for customers to take a risk on their technology or a cost model (leveraging overseas resources) to make it worth the risk.
And then founders need just the right amount of hunger and hubris to think they can succeed.
As a Ph.D. student, “you’re used to living on $2,000 a month…so when your salary gets raised to $3,000 a month you think, ‘Wow it’s a raise!’” he said. “You (also) have this unfounded confidence that you can fight against incumbents in the world — $10 billion companies. It’s a very good time to start a company.”
Hello, Elon, Is that You?
The team had secured $150,000 grant from the National Science Foundation in 2013 and then, over time, more from angel investors including the legendary microprocessor designer turned investor Dan Dobberpuhl and then $10 million in Series A money from Walden Riverwood and Atlantic Bridge Ventures.
But how they got there is fascinating: The co-founders pulled together a list of possible investors and cold-called them. Elon Musk, Tesla’s founder, didn’t respond, but many others did, including SanDisk founder Eli Harari.
“We learned to try to stay persistent even when there were naysayers,” Kim said.
Having used Virtuoso software in college, they approached Cadence about additional EDA tools support.
“They had so much passion and enthusiasm,” said Evelyn Puchta, Cadence account manager for startups. “I was so taken with them and we did everything to help them get started. It’s so awesome to hear Kim tell us they closed $10 million in funding.”
Along the way, they learned valuable lessons and did some important pivoting--the biggest of which was a market switch. Lion Semiconductor initially pushed its technology toward FPGAs and enterprise systems because their technology could not support mobile specifications. In the mobile market, both the input and output voltages of the voltage converter have to be very wide range. The battery voltage has a wide range and the output to the processor has to be wide range because of dynamic voltage scaling, Kim said.
But Brian Long, with Atlantic Ventures, pushed them hard toward mobile because of the market opportunities, he said.
So they went back to the drawing board and came up with a new DC-DC design that could maintain high efficiencies across such stringent mobile specifications.
“We patented some technology and met with customers. They liked the idea,” Kim said.
There’s still a long road ahead of Kim and his colleagues at Lion, based in San Francisco. But he sounds like an old semiconductor hand when he looks ahead:
“We don’t have an exit strategy and we try not to think about it. It messes up your brain. If we do our job, if we get a design win and profitability, then good exits will follow.”
—Should More Semiconductor, EDA Startups Look to Kickstarter?