SM: You have been doing this for two years, where are you in terms of products?
FW: We don’t produce the DIMM, we produce controller chips. We sell controller chips to DIMM makers. Our partners today are Smart Modular Technologies and Hynix. They take their DRAM, put it on a board with our chips to make a DIMM, and sell it to their normal customers.
SM: Who are the end customers using them?
FW: It is a complicated supply chain. We sell our chips to Hynix and others who in turn sell them to OEMs. We announced that Verari, Colfax, Rackable Systems, and Appro are using our DIMMs and sell them to their high performance computing customers.
SM: Does Sun and SGI play in this market?
FW: They do, as well as all of the major OEMs. They are all aware of what we do. We do not have anything public with any major OEM, but we certainly intend to. They are aware of us and I think we have a good value proposition for them.
SM: Where do you see this going and what is timeframe? What are your next big milestones?
FW: We need to sell a lot more. That is a matter of filling the pipeline. We have to get the customer comfortable at each stage and qualify the product in their systems. This year is mostly about getting that supply chain in place and getting some high profile customer wins. We were at the Intel developer forum in Shanghai. We showed the DDR2 product running in Intel’s new San Clemente platform which is a very nice platform based on a new chipset which runs with registered DIMMS rather than buffered DIMMS. They put two quad core processors on a blade form factor which makes an incredible machine with small form factor, but has only 6 DIMM slots. Our DIMMS can take that system up to 48GB and make a really wonderful computer.
Perhaps more exciting for the future we had the DDR3 version up and running on Intel’s upcoming DDR3 Memory Development Platform. We showed our DIMMS working inside of Intel’s next generation platform. That was exciting and proves our DDR3 capability. In 2009 it will be about the DDR3 ramp of our product.
SM: Where is AMD at with all of this? I imagine you know there problems, are you a solution?
FW: Our DDR2 product, which I just described, working in the Intel platform, was really designed for the AMD platforms. Most of our early opportunities are in the AMD platforms with the Barcelona processor. Those platforms have a lot of DIMM slots. In our lab we have quad socket AMD machines with 32 DIMMs. With our DIMMs installed we have a quarter terabyte of memory in a 1U form factor. Sun has a very high end machine with 8 Opteron processors. It has 64 DIMM slots, which we have filled to provide a half terabyte of memory in one computer.
SM: How does the memory industry view you?
FW: We make it possible to sell more DRAM into the server market. That is a good thing because the margins in the client market are really bad for DRAM. They are better in the server market. In general we are good and liked by the memory industry for creating that capability.
There is some level at which we create competition against next generation DRAMs. If you compare the DIMM we can make with 1GB DRAM to what industry can do with 2GB next generation DRAM, they are semi-equivalent products. Once industry accepts us then we can use their 2GB DRAM as a platform to create even larger DIMMs. There will be a short transition period where industry will adjust to our presence, but in the long run we allow them to sell more of their product into the higher margin businesses.
SM: What is the market size you are going for and where does it fall in the food chain?
FW: It is the higher end where we are not looking at tens of millions of DIMMS. The market has very good margins When we started the company the projections of the analysts were that 8GB DIMMs were going to be $4,000 this year. Our DIMMs are selling at list prices of $1500 or less. That creates some demand.
Over time as the DRAMs get cheaper and the volumes go up, we think our technology can be applied into the $500 product range. There is an infinite demand for memory. For a high end workstation, spending $20,000 is fine, but you are not going to spend that for 100 workstations. Right now we are high end workstations and servers, in the $20,000 to $100,000 range replacing the functionality of machines in the $1M range.