TSMC holds lessons in success, rising from its first access to Philips' process technology to building the next Apple iPhone SoC.
You probably heard that TSMC recently won a big commitment from Apple for volume production of the A8 processor. That's just one significant data point for the foundry's strong market position and an indication of its future success.
Market researchers say TSMC supplies as much as 80 percent of the world's 28nm silicon and will earn about $6 billion from the node in 2013. But it wasn't always that way.
Do you remember the Bay Area earthquake on October 17, 1989? I do, because I had just completed my 14th interview at VLSI Technology and my hiring manager, Cliff Roe, had told me he was preparing an offer for me to join what I considered the best cell-based IC vendor in the world.
After a few days of clean up and recovery after the quake, I got my offer and joined VLSI as Philips' alliance manager. I already knew about Philips' efforts to ramp up a CMOS process at a company in Taiwan. That kickstarted TSMC, but I didn't realize its full significance.
In 1995, I was VLSI's marketing director for consumer ASICs and ASSPs and heard many times from my customers: "Sorry, I can't give VLSI this order, TSMC offers a much lower price. This frustrating experience and the decline of VLSI's in-house EDA tools (developed by Compass) motivated me to join an EDA company, ViewLogic.
I didn't forget what TSMC did to my business at VLSI. I simply thought, if you can't beat them, join them!
In 1997, one of my first customer visits was to TSMC's office on Technology Drive near the San Jose Airport where they worked hard to convince US customers to buy wafers from their Taiwan fab. Their timing was perfect, because the traditional ASIC business was giving way to Customer-owned Tooling. IC designers could submit GDSII streams to their manufacturing partners. ViewLogic’s management realized this trend and allowed me to hire a dedicated TSMC alliance manager.
Before the year was over, Synopsys decided to acquire ViewLogic. My new manager didn't support my plan to form an alliance with TSMC, but I kept the project alive as a skunkworks.
Soon after joining Synopsys I visited Cliff Hou and others at TSMC in Taiwan and discussed their needs for an EDA reference flow and how Synopsys should support them. It took about a year until my skunkworks project smelled like roses at Synopsys.
Eventually management allowed me to hire an engineer to work at TSMC's facility in Taiwan. In 1999, we developed a reference flow to design a 1 million gate chip using Design Compiler and PrimeTime. It encouraged the rapidly growing fabless IC vendor community to base internal design flows on Synopsys tools.
Fast forward to October 1, 2013. I attended TSMC's OIP Symposium and was very impressed. More than 1,000 attendees wanted to hear a technology update from TSMC's CTO, Jack Sun, and VP of design enablement, Cliff Hou.
The young engineer I met 15 years ago is now responsible for an entire division. Cliff not only directs hundreds of TSMC engineers, he influences an army of partners in EDA, libraries, IP development, and design services. He even works in my new focus area of 2.5 and 3D-ICs.
It was great to hear at the event about what TSMC is doing in Wide I/O and High-Bandwidth Memory and the foundry's commitment in general to vertical die stacking using Cadence and Mentor tools.
In TSMC's rise, I see some general lessons for building a successful company:
- Refine your vision by analyzing market trends and requirements
- Plan your market entry and expansion carefully
- Hire and promote excellent people
- Execute, execute, execute
- Team up with key partners who complement your capabilities
- Keep your costs down
- Show integrity
- Consider the long-term impact of every step
And yes, strong financial support is also important, especially at the beginning.