Open Ambition

Tales from the early-ish days of Silicon Valley

By Peter Zaballos

For the first 2+ years of my tenure at LSI Logic I was the product marketing manager for Northern California and the Pacific Northwest. I didn’t ask for this geography, it was the job they offered me. I took it without fully appreciating what I was about to get into, and accepted it largely on the basis of the quality of the people I knew there, and met during my interviews. 

All of them were super, super smart. Ambitious. Uncompromising. Kind and fair.

I had spent the past year (to the day) as a Product Marketing Engineer at Fairchild Semiconductor, responsible for their high performance ECL (emitter coupled logic) product line. And a colleague there got recruited to LSI. Within a month or so, I was also recruited to join the company. In 1983. I believe I was employee #87.

The co-founder and CEO of LSI Logic had been the CEO of Fairchild, hence there were lots of Fairchild folks jumping ship to join him.

[BTW, Fairchild Semiconductor was formed when the traitorous eight fled Schockley Semiconductor. Among those eight: Bob Noyce and Gordon Moore (founders of Intel) and Eugene Kleiner, co-founder of the legendary VC firm, Keliner Perkins). Fairchild is truly the foundation of what became Silicon Valley]

All I knew about LSI Logic was the reputation it had for creating a new category of semiconductors. Semiconductors that could be easily and rapidly customized. Back in the early 1980s it could take two years to create a custom semiconductor, because the entire chip had to be designed from top to bottom. And computer simulation tools were still in their infancy, so a fully custom chip pretty much had to be designed by hand.

Semiconductors are sort of like a pizza. There’s a silicon substrate, and on top of that different chemicals are added to different portions of the chip to affect the conductivity (or resistivity) of the silicon. And then layers of insulators are added, that are put down in such a way to keep openings for layers of metal to be added, connecting the spots needed for electricity to run.

So a fully custom semiconductor required that everything – all the way down to the substrate – had to be designed and built. And since ultimately the cost of a chip is proportional to the amount of silicon it takes up, a lot of work went into laying the chip out in such a way as to pack everything together as tightly as possible. Custom chips were incredibly expensive and time consuming to build. So they were really only used for highly specialized and valuable products – that could put up with a two year development timeline.

The founders of LSI Logic – Wilf Corrigan, Rob Walker, Bill O’Meara and Mick Bohn – saw a huge opportunity to disrupt this market. Making it possible to create a customized semiconductor in less than twelve weeks. Yes, twelve weeks.

The first innovation was in separating the semiconductor fabrication process into two parts – creating a standardized substrate and then a customized metalization layer on top of that. Rather than starting with raw silicon, LSI started with a large array of transistor “gates” – initially 1,000s of these, later millions. These base wafers could be mass produced by contracted semiconductor fabs and sent to LSI’s facility in Milpitas. And in the Milpitas facility we could connect those gates in the metalization process to create the specific functionality the customer designed.

Let’s pause for a bit to talk about how semiconductors chips are made. With a base layer of transistors laid down, those transistors can be connected to each other by laying down metal – in very thin lines – across the chip. Which means that you need a pattern of where the metal should go, as well as a pattern where you need to insulate transistors used for other parts of the design from the ones you want to connect.

The result is a series of masking steps that are performed. You might produce a masking pattern of insulation, followed by a different masking pattern of metal connections, and then add another insulation masking pattern on top of that protecting and exposing different connections, and add another layer of connections. The set of “masks” would be the recipe to create the functions of the chip you were fabricating.

So by separately creating the base layer of transistor gate arrays there was no need to as tightly pack the transistors together. You gave up some real estate for overall speed and flexibility of production. And the real value came in the process of connecting all those transistors together to perform the specific function the customer needed.

The second innovation was creating powerful semiconductor design and simulation software (running on IBM 370 mainframes) that would enable a customer to take a logical design of the custom product they wanted to build and simulate its performance. Up until this point companies had to build a physical prototype of a product using a collection of building block, general purpose chips to see if the product would work. And then begin the process of converting that into a custom chip. One of the factors driving the 2+ year custom chip timelines.

LSI’s third innovation was that the computer simulation could produce the connection pattern for those base wafers – to produce the design the customer wanted – and create all the “mask” information needed for a mask vendor to create all the fabrication templates needed to build that custom chip. 

Once the mask set was sent back to LSI, then our fab in Milpitas (which only needed to manufacture the connection pattern) could produce a set of prototypes in about 10-12 weeks.

If the customer had a well thought through design when they showed up at our simulation center, they could have that design simulated and validated in a few weeks. All told, from showing up at our facility to having prototype chips in hand could be as little as 3-4 months.

This was earth shattering for a number of reasons. First was simply shrinking that time to prototypes. From years to months. Second was the economics. Because we were typically taking designs that would have required tens or hundreds of discrete semiconductors and turning all of that into one chip, our products would pay for themselves in cost savings to the customer. And the resulting product could be dramatically smaller, physically. So entire new categories of products became possible. The IBM PC, the Macintosh, disk drives, network routers, Sun Microsystem workstations and servers. Eventually cell phones and digital audio and video.

But even more important than economics, what we did at LSI Logic was disrupt how rapidly a company could bring a product to market. And speed up the tempo of innovation in a market. Forget about cost savings. The value of dramatically speeding the pace of innovation is what transformed the market.

Going back to taking that offer to join LSI. My territory was Northern California and the Pacific Northwest. My customers – who were startups back then – were Sun Microsystems, Seagate, Silicon Graphics, Apple, Cisco. We put those companies in the positions they soon commanded, to transform their industries through the rapid pace of their own innovation.

And when these customers came to us with their new product designs, delivering the resulting prototypes became a treasured ritual. I remember walking into Silicon Graphics with the prototypes that would power their first workstation. And later seeing them boot up the system and load Flight Simulator – and the tens of people crowded around the engineer flying an airplane on a computer screen.

So, what LSI Logic made possible was taking the functions that would fill a computer the size of a refrigerator and shrinking that onto a single chip. And making it possible for that single chip to go from design to prototypes in ~12 weeks.

We fundamentally changed the economics of computing and the pace of innovation. 

You can listen to Wilf Corrigan tell the origin story of LSI Logic in this recording of his keynote from our 1986 Sales Conference, which was held in Sunnyvale. I was in the room for this, and it is just as thrilling today as it was back then. And Wilf does an excellent job highlighting the many talented folks who made our business and technology so successful.

And I thought that everybody got a job like this out of college. I mean I didn’t exactly look for this job. Someone I worked with at Fairchild called me and asked me to join him there. I came over for an afternoon of interviews and was offered the job. Responsible for perhaps the greatest single concentration of computing innovation in the world. 

It is only with hindsight that I can appreciate how fortunate I was. It’s also where I now see the pattern that fueled and informed my career in high growth technology companies: working with highly talented, highly ethical people. And those people tend to congregate with other good people.

Steve Jobs did a wonderful job explaining the impossibility of predicting career paths in his Stanford Commencement speech – that you can only connect the dots of your career path in hindsight. Joining LSI Logic fundamentally shaped the path of my career in a few ways.

First, it taught me how important it is that you work in a company where the definition of “good” is far, far greater than what other people experience. “Good” at LSI Logic was making it possible for Cisco to deliver its first router. For Seagate to ship its first hard drives. For Sun Microsystems to ship its first SPARC workstation.

“Good” meant that you delivered a product that was 10x what the customer was expecting or even imagined. LSI Logic was at the forefront of that “10x, not 10%” mentality that causes a business to create an entirely new category.

And to do that, the company needed embrace another Jobs-ian tenet: “A players hire A players, B players hire C players.” Bill O’Meara ended up taking over as CEO of C-Cube Microsystems and enabling digital film and digital TV (and brought a lot of us LSI folks along with him). My colleagues JenSen Huang founded Nvidia, Bill Tai became a world class venture capitalist, David Baillie became CEO of not one, but two technology companies, and John Daane became CEO of Altera.

We all left our tenure at LSI Logic clearly understanding what it took to build a business that invented an entirely new class of computing technologies, what it meant to disrupt a market, and how to fundamentally change what your customers thought was possible.

I left LSI Logic after nearly seven years there to go to business school, but shortly after graduating and right after getting married, I got a call from Bill O’Meara, telling me about a company he had just become the CEO of – C-Cube Microsystems – and asked me to join him there. And the next adventure unfolded.