by Debra Vogler, Senior Technical Editor
With the growing trend to form cluster alliances as IDMs accelerate a fab-lite/process-lite business model, one might ask how the design, tape-out, and responsibility for ensuring a manufactured product actually emerges with a profitable yield. Global Unichip Corp. (GUC) seems to have arrived on the scene with an answer: “virtually vertical integration” provided by a design service foundry.
Founded in 1998 at Taiwan’s Hsinchu Science Park, GUC entered into a partnership with TSMC in 2003, and this key shareholder and sole foundry partner remains a strong influence. Current GUC board members include chairman FC Tseng (TSMC’s vice chairman), president & COO Jim Lai (former TSMC USA director), and Lora Ho (TSMC’s CFO). KC Shih, founder, vice chairman, and CEO of GUC, was also a founder of Faraday Technology Corp.
The company offers design services and will take a design all the way to tape-out, production, and beyond, according to Lai. “We have no intention of being a fabless company,” he told WaferNEWS. “We own IP but not products.” Turnkey services include package engineering, test engineering, product engineering, quality and reliability, and logistics. The company’s revenue doubled in 2006 to $103 million.
Like many fabless and design services firms, GUC has had to identify its market sweet spot, and has targeted advanced technology ? 130nm, 90nm, 65nm, 45nm, etc. “We’re not forgetting about the other nodes, but to start out, we are focusing on the top of the pyramid,” Lai said. “We can differentiate ourselves better among competitors, and the need [for a design foundry] will be more obvious at the more advanced technology nodes. The revenue can prove that this strategy works very well.”
Pointing out that NRE is usually a leading indicator of future revenue, Lai noted that since 2004 the company has completed 45 tapeouts at 130nm; ~12 at 90nm; and 3 at 65nm, which were done earlier this year. “Those tape-outs will eventually turn into turnkey revenue in 2008 and 2009,” he said. “We will start to see 90nm production revenue this year. Most of the 90nm revenue we did last year was from NRE and from tape-outs ($5-6 million was in 90nm).” In terms of sales breakdown by region: most of GUC’s revenue is from the US and Japan, but they are trying to make more inroads into Europe and Korea.
Lai noted that GUC does play into the vision of “cluster alliances,” as recently proposed by ASE Group’s Tien Wu, but with a different label. “We call it virtually vertical integration,” he said. Being a virtual IDM (i.e., a design foundry) means the company is involved in all aspects of the food chain for a particular IC product, but with so many parts of the design and manufacturing processes involved, including packaging and test, “There’s no more design for multiple sources or foundries ? you want to optimize your design to a particular foundry/process of your own,” explained company CEO KC Shih. “We have to involve the entire food chain to be able to come up with the most optimal design — so that’s why we call ourselves a virtual IDM. We’re doing the virtual integration into a single chip without the burden of the capital investment of an IDM — not even the entire suite of IP, because we can use either external IP or internal IP.”
Shih further added that the company likes to consider itself the bridge between the fab-lite IDM and the foundry. “We believe we are in a good position to help those IDMs that have made the decision to go fab-lite/process-lite, but the trend is just starting.”
With the process-lite/fab-lite model gaining traction in the industry, Lai introduces another twist to the equation, calling his company’s model a “tool-lite” approach. “People that come to us don’t have to buy all the tools [that service maskmaking, litho-aware DFM, etc.], they can just outsource that part of the design to us,” he said.
This tool-lite concept begs a significant question: How will EDA companies compete and survive once foundry design companies take over? Lai doesn’t think this will have an impact, because the smaller fabless companies didn’t buy the high-end tools anyway — and these smaller design houses will still need the front-end tools such as synthesis and verification. — D.V