On the occasion of a recent business trip to Taipei, DigiTimes had an opportunity to talk with Randy Levine Ph.D, president and CEO of startup company ZettaCore, Inc., and Craig Rhodine, VP of engineering at ZettaCore. Spun out of academic research at the University of California and North Carolina State University, ZettaCore is a specialist in molecular memory.
Headquartered in Denver, Colorado, ZettaCore received its first round of funding in December of 2001 and a second round in December of 2003. To date, the company has raised US$23 million.
This is part one of a three-part interview. The second part will follow on Thursday, September 30, and the third part will follow on Friday, October 1.
DigiTimes: What stage are you at with the technology? Are you still in R&D or do you now have actual products or product samples? Has ZettaCore’s molecular memory been demoed in real-life devices?
ZettaCore: We have built working prototypes on the scale of a megabit. These are cost-effective for us but large enough to impress people. We are in the process of characterizing the technology in a real-life device and optimizing the process parameters for manufacturing.
We made two choices early on, both very important. One was to develop a technology that implemented memory functions in the same way they are implemented now, which is by storing charge. That is what people already know how to do. That's important because the same basic design techniques carry through; the same test techniques carry through.
The other decision we made, which was really a business decision, was to implement the technology first in a way that was as compatible as possible with standard CMOS processes because we wanted to make sure that people could adopt this technology as easily as possible. There's already enough of a barrier to adopt anything different in this industry. To bring in completely different processes for a completely different design was something we didn't think was wise, although there will be improvements to the technology later. We think it is more appropriate to introduce those changes after people have gained some confidence in the technology, and it has proven itself.
To date, we are satisfied with the performance of the technology. The next step will be to get to production-scale devices.
DigiTimes: Do you have a timeframe for commercialization?
ZettaCore: We get asked that a lot and we try not to commit to anything specific. You never know what kind of obstacles you are going to hit when you try real production in a real fab. Second, it's going to depend somewhat on specific customers we are working with and their design and product cycles. We have commercial partnerships, but we don’t disclose who they are. We will be in a position to finish our characterization later this year and begin production of real products soon after that. That we can say. Now how long does that take to get to market? That's where we don't want to set any specific expectations.
DigiTimes: We understand you are a fabless design company, so who is or will be fabbing your memories? Are you in Taiwan to talk to the two major foundries, for example?
ZettaCore: We are in Taiwan to talk to potential customers or licensees of our tech but we are careful not disclose specifics of discussions.
Let me just say a few words about business models. There are three ways we'll get paid. One way is to partner with people as in effect a fabless semiconductor company.
We'll also license our technology for certain applications. The best example of that is embedded memory. It's a very good technology for embedded applications.
The third way we'll get paid is by selling molecules. This is a technology that uses specially designed molecules to store information, and we will make and sell those molecules.
Manufacturing the molecular material is a chemical manufacturing process, and that is something we can afford to do on our own and then sell at a profit. On the other hand, we are not in a position to build our own fab. We are not planning to do that.
DigiTimes: What facilities would be required for producing the molecular material?
ZettaCore: A very modest-size industrial chemistry lab. We already have the staffing that would allow us to support our first customers. To do more than that would only take a couple of dozen people, no more. Chemical processing is not a labor- or capital-intensive activity.
DigiTimes: What kind of process is involved? Is your type of molecular memory difficult or easy to fab, or to put it another way, is it expensive to fab or about the same cost as other memory chips?
ZettaCore: One of the big advantages of this technology is that it is much less costly than other memory-chip technologies, certainly conventional memory technologies. The reason is very simple. To build a conventional memory cell – a volatile memory cell – you have to build a trench or a stack. With our technology you don't have to do that. All the fab steps and materials you use to make a trench or a stack are not required, so the cost goes down tremendously.
For an embedded memory, this type of molecular memory can be made on a conventional logic line, without significant additional process risk, so it offers a density improvement over embedded SRAM, the predominant embedded memory currently. In that case the savings are even greater because for the same cost you can create between five and 10 times as much memory as you can now. Said another way, for the same area on the chip you can put five to 10 times as much memory on the chip.
This is part one of a three-part interview. The second part will follow on Thursday the 30, and the third part will follow on Friday, October 1.
Article edited by John McClure
