A few weeks ago, we were surprised to run into a new launch of a semiconductor startup. Semiconductor startups seem to be quite rare in today's environment, with few companies being funded or launched, due both to the difficulty of bringing products to market, lack of investment capital, as well as lots of established, large companies in the industry. Irvine-based RFaxis (www.rfaxis.com) is tackling the RF front end market, and we spoke with Mike Neshat, President and CEO of the company, to learn more about where the firm fits, why Mike decided to start a chip firm, and how far along it is now with its product.
Thanks for taking the time to speak to us. For starters, let's talk about your technology. What part of the wireless stack are you tackling?
Mike Neshat: We founded the company almost two years ago, and we're actually not as new as people think. We've kept under the radar, primarily because for the first twenty months of the company's life, I funded the company myself. We're now funded by a bunch of angels. Our technology we have just introduced, is the first and only single chip, single die BiCMOS front end in the world. There is no other product like it out there. In terms of markets, we address the Zigbee, Bluetooth, and 2.4 Ghz wireless market. We have future product which will tape out in about seven weeks--five or six more products--which address the WiMax market, dual-band, 2.45 Ghz, and 5 Ghz market, aimed at cell phones and BlueTooth. The goal is to capture as much of the front end sockets out there as we can.
How much of an advantage does your chipset offer over existing technology, and how does it compare over existing levels of integration?
Mike Neshat: Today, there is no single chip, single die solution out there. The highest level of integration--which you might find from RFMD, or Skyworks, or SiGe Semiconductor--is the front end module. They've taken a bunch of Gallium Arsenide (GaAs) components and put them together in a single package. But, they've pretty much reached the end of their packaging technology. With front end modules, there are still external components for matching, etc. In our case, we have put everything on a single die, the PA, LNA, matching, switching, and harmonic filter. It's plug-an-play, and can be dropped in with just a bypass capacitor. The way we've been describing this to people has been, you can treat our RF device like a digital chip--all of the RF design is out of the equation.
Is there any cost difference and advantage of this over existing solutions?
Mike Neshat: We were just at Computex in Taiwan, and our pricing is definitely, significantly lower than front end modules, and even lower in cost than how you'd do this traditionally. In terms of pricing, no one can compete. We can also sell it aggressively, at a decent price, and make a health profit margin.
How far along are you to production shipments of your chips?
Mike Neshat: We introduced two products at Computex. One, which we refer to as the RF2401, is intended for the Zigbee and Bluetooth markets. That product will be in production in Q4, probably October. It's the same part that came out as our first engineering sample. The part tested so well, we are taking that engineering sample to production without any enhancements to our die. Our RF2402 product, our wireless LAN product, will be in production in Q2 of 2010.
Developing a chip has been very expensive--have prices gone down now, or how were you able to self-fund the effort?
Mike Neshat: I'm broke now. (chuckles). Chip development costs have definitely not gone down. Good RF engineers, as you know, are quite expensive, plus you have to spend on tools, spend money on the fab for what you have. The price hasn't gone down, but I chose to fund it myself initially and then went to angels, because I felt the right time to go to VCs was when the valuation was higher. It has been a very expensive ordeal.
RF experience isn't easy to find--what's your background, and where's your team from?
Mike Neshat: I've been in the industry for 26 years, and have been involved in several startups, which were successfully acquired. For the most part, for the past 20 years, I've been on the marketing and sales side for companies like Rockwell, GEC Plessey Semiconductor, Silicon Systems, Fresco Microchip, Synad Technologies, WyLinx, and others. I've been doing this for a long time. Oleksandr Gorbachov, my CTO, has been in wireless microwave and millimeter wave technology for 27 or 28 years. He's been at companies like Synad technologies, Hexawave, Gatax Technology, and at the ORION Institute in Kiev. Combined between the two of us, we've been at this for 54 years.
There is lots of competition from the big companies, legal threats in the industry for chip startups to navigate today, how are you handling that?
Mike Neshat: The good news is, we don't compete with Broadcom, Cirrus, or any of the big guys. If anything, we complement them really well. We're not in the transceiver, or broadband business, and are strictly RF front end. Anything we'll be doing will have reference designs with Broadcom and Atheros and the others. So there's no legal issues for us, and if they don't infringe on our patents, there's no worry for them. We've done something revolutionary--no one has been able to do this in BiCMOS. Traditionally, all of these RF front end solutions have been in Gallium Arsenide, and our patents protect us no matter what technology we use. We're in good shape, and are excited, and think we'll probably be acquired by one of the large companies in a few years for a lot of money.
If you don't mind answering, what fab are you using?
Mike Neshat: Our foundry partner is STMicro. We're not using an exotic fab or process. It's standard a BiCMOS, 7 RF quarter-micron process. We're also looking at using TSMC as a second source. That's the nice thing about our technology--it's not exotic, it works on any mature, proven, BiCMOS. We have 14-16 patents filed, and our goal is to have 35 or so patents filed by the end of the year. What we have patented it stricly having to do with our architecture, which is truly independent of the kind of process.