San Francisco A startup will surface at Semicon West here today with a portfolio of technologies that it says can speed chip designs and processes from lab to fab, reining in runaway R&D costs and even allowing the function to be outsourced. The approach is just one of a host of initiatives being pitched to remake the chip industry's research-and-development model.
Intermolecular Inc. (San Jose, Calif.) says its High-Productivity Combinatorial (HPC) platform of "fab in a lab" technologies will facilitate R&D of IC materials, processes and device structures. The startup is offering a trio of programs.
One is a line of tools that brings the company into the semiconductor equipment business. The systems are not meant for the production line; rather, they are 300-mm lab tools for R&D. The machines can be configured to support many critical fab processing steps--such as cleaning, electroless deposition and surface preparation--in a modular prototype system.
Intermolecular is also entering the intellectual-property arena, a move that could put the startup on a collision course with the likes of Applied Materials, Lam Research, Novellus and Tokyo Electron Ltd. One of the startup's first IP offerings is a "molecular masking layer" for use in capping-layer applications at the 32-nanometer node.
Finally, the company can develop processes and technologies for customers within its own clean-room facilities, said Intermolecular CEO David Lazovsky.
"New [R&D] processes and integration methods are required that could take several years" for the industry to develop, Lazovsky said. HPC "speeds R&D learning rates by orders of magnitude compared with conventional R&D methods."
Lazovsky insisted the startup is not competing against IC makers, tool vendors or R&D consortia such as Albany Nanotech, IMEC, Selete and Sematech. "We can work with them," he said.
Intermolecular is addressing what he called the "perfect storm" in semiconductor R&D. "R&D spending is running out of control," he said. "Our mission is to make R&D more productive."
While it remains to be seen whether Intermolecular's strategy will work, there is a crying need in the chip industry for feasible new R&D models.
"R&D budgets are being honed," said Robert Lineback, an analyst with IC Insights Inc. (Scottsdale, Ariz.). "More and more basic manufacturing technologies are being moved to consortia, universities and partnerships.
"It's still too early to know if these partnerships will hold together well enough to succeed at the 32-nm node and beyond. But the pressure is much greater to make the partnerships work, since product designs can't be delayed, or market windows will be missed."
Compounding the problem is the runaway cost of R&D, especially as a percentage of revenue. In 1978, total semiconductor R&D was $600 million, according to IC Insights. Intermolecular claims that figure had grown to $45 billion by 2006 and is expected to hit a whopping $100 billion in 2012.
Semiconductor revenue is projected to grow at a compound annual rate of 6.5 percent from 2004 to 2020, according to VLSI Research Inc. (Santa Clara, Calif.). But R&D spending will rise 12.2 percent on average in the same time frame, the research firm predicts.
IC design, process and equipment costs are the main culprits in R&D's rising tally. Process technology R&D costs alone are expected to jump from $2.4 billion at the 45-nm node to $3 billion at 32 nm, according to analysts.
For years, many chip makers handled the bulk of their R&D efforts in-house. Starting in the mid-1980s, R&D consortia were launched to lessen the burden.
Other R&D models have recently emerged. IBM Corp. formed a technology alliance with various chip makers to offset R&D costs. More recently, silicon foundries have jumped into R&D.
Now, specialty R&D service providers are entering the market. In June, SVTC Technologies Inc. (San Jose, Calif.) debuted after being spun off from Cypress Semiconductor Inc. SVTC has a 200-mm prototype plant and has rolled out a range of foundry-like services for lab-to-fab development of product ideas.
Also last month, three companies merged to form Nano Integrated Solutions Inc. (Santa Clara, Calif.). The new venture intends to provide a one-stop shop of services, including testing, failure analysis, reliability qualification and board design.