Ottawa can get pretty cold at this time of year, but Jim Roche says Tundra has never been hotter.
December marks the 10th anniversary of Tundra Semiconductor, which was founded by former employees of Newbridge Networks (later purchased by Alcatel). The company carved out a place for itself by focusing on the technology used to connect all the components and sub-systems in almost any embedded computer. It used the term “system interconnect” to refer to single-board products that interface with CPUs, memory, I/O and other complex parts of IT. According to some analysts, a good half of Tundra’s product line has no clear competition, but as system design gets more complex, Roche, its CEO, doesn’t expect things to stay that way for long.
“There’s always a tinge of regret that our little niche is now attracting competition,” he said, adding that a decade of experience gives Tundra a major advantage over new entrants. “We’re way ahead of these guys in terms of the core technologies, the patents and IP we’ve got, the customers and relationships in the market.”
Tundra has been a quiet Canadian success story, though it got a strong endorsement from the government in September when Technology Partnerships Canada agreed to invest $7 million into its R&D efforts for next-generation RapidIO System Interconnect products. Roche spoke with ITBusiness.ca on Wednesday about the lessons learned in the first 10 years and his plans for the next.
ITBusiness.ca: Take us back to the early days of Tundra’s history. When did it make sense to spin off from Newbridge?
Jim Roche: When I graduated from university I got a job offer from Terry Matthews. That was back in 1986 – I was one of the first employees at Newbridge. I remember when I walked in the front door, I think there were four people in the company at that time. I looked around and I thought, “Boy, Terry’s leaving too much space. We need to do something about that.” (laughs) I had the experience of working with Terry and the rest of the Newbridge crew as we grew that company from nothing to a sizable, very successful international high-tech company. My role at Newbridge in those days was in engineering and management, and eventually I moved within Newbridge to run one of the divisions. I wanted to experience the early days of structuring, building and growing a company from an executive posititon. That was one of the key bits of my motivation for starting Tundra Semiconductor.
In 1995 we acquired some of the assets from one of the divisions of Newbridge at that time, and remained close to Newbridge, then weaned ourselves away. Terry had started up what he called the affiliate program, where a number of companies that were satellite companies of Newbridge developed a closer and closer relationship with Newbridge. We had the opposite experience of being an affiliate that was on its way away from Newbridge, as we established ourselves as an organization.
ITB: How is the competitive landscape different today than it was 10 years ago?
JR: In 1995, this was before the tech bubble had really taken off, but still high-tech was ramping up. The environment was, I think, every conducive to starting an organization. We took the company public in about three years – three years and about two months, I guess – which is a very short ramp time to go from a cold start to public. That was because we had great technology at the time but also because of the tech boom. Then of course in 2000, things went haywire. Every time I looked out my window I’d see a Porcshe going up and down the road as some other high-tech executive went and cashed in some options. Then in 2000-2001, the bubble burst and the environment completely changed. It’s been five years since then and at Tundra, our revenues dropped to less than a third of what we had been running at in the tech bubble within six months. We’ve been recovering and growing since then.
ITB: In some respects Tundra has defined the system interconnect market. When did you recognize that as an opportunity that hadn’t been pursued?
JR: We all had a lot of experience in very complex, high-performance communications systems. We knew that what was important was the technology that glued the bits and pieces together. What we anticipated was that the interconnect would have to run at higher and higher speeds as time went on and this was going to put increasing pressure on the designers at companies like Cisco, Ericsson, Hewlett-Packard and EMC. We thought what would happen as a result of that was that these companies would be looking for off-the-shelf interconnect solutions. We anticipated an outsourcing trend in this type of technology.
ITB:To what extent is the plethora of devices and the mobile enterprise growing that market opportunity for you now?
JR: The system interconnect problem is something that exists in just about any mid- to high-performance system. What’s happening is as time goes on, every system is driving data faster, and it’s driving more complex forms of data around the system. It turns out that there’s a hard stop in terms of how fast the old style of interconnect, the bus, can go. To go past that you have to go to something else, which is switched fabric architecture. The mid-to-low performance systems are hitting that hard stop in bus performance, and that’s forcing them to go to a completely new architecture for interconnect, and we’re in the middle of that. But it’s not being driven by any one technology, like enterprise mobility. That is an application that we play in, but we’re also very active in core networking, IP networking, voice-over-IP, in 3G networks, in WiMax. We’re very active in printing applications, industrial control applications. Every one of these kinds of applications has a need to move data around the system really, really fast.
ITB: A fair proportion of companies still handle system interconnect themselves. What’s driving the transition around outsourcing?
JR: Yeah, it’s still about 90 per cent of companies that are designing that internally, though it’s hard to measure. To go to Cisco and say, “How much money do you spend on system interconnect, and what percentage of it is internal and external,” they don’t release that kind of data. About 90 per cent of it is still insourced. What’s driving the outsourcing is the incremental cost of developing this stuff. If you look at the semiconductor industry, to design chips 10 years ago would cost us a few hundred thousand dollars to get something into production. It was relatively inexpensive. Today, you can’t get a chip into production for less than about $10 million. It’s very, very expensive and as you move to smaller geometries – to 90 nanometres and 65 nanometres and beyond – that cost is growing exponentially. Economically, it only makes sense to invest in this sort of technology if you’ve got significant volume or you’ve got a real significant value proposition in the intellectual property.
Let’s take a look at a company like Cisco. Why would they do their own system interconnect? Right now, they have the volume to justify the cost. But as the costs rise, that economic model starts to shift, and they decide it doesn’t make sense to invest that money on a system interconnect solution, and they go out of house. The second thing that happens is for companies like Cisco, when they get to that point, the next question they ask themselves is, “Does this system interconnect give us a competitive differentiator?” In other words, “Does this make us better than Juniper?” And the answer, quite frankly, is usually not really. The reason is that the system interconnect, while it’s critically important to these products, isn’t of itself the differentiator. The differentiator tends to exist in other silicon, or more often than not, software. There was a time when companies like Cisco and Nortel did their own processors. Almost no one does that anymore. The same thing is happening with system interconnect. It’s just happening 15 to 20 years later.