November 01, 2010

Save our Semiconductors

By: by Ken Stier, Time Magazine


Nestled in a red-pine forest 25 miles (40 km) north of Albany, N.Y., close to where Wernher von Braun experimented with rocket fuels, a new $5.5 billion computer-chip-fabrication plant--deploying physics far more sophisticated than that used in rockets--is being built. The fab will be a critical locomotive for the local economy. It is also good news for the U.S. semiconductor industry, which is the primary fount of the economy's productivity gains--and all the gadgetry we consider critical to modern life.

The fab is being built by GlobalFoundries, a joint venture led by Advanced Micro Devices (AMD), which spun off its manufacturing arm to focus on chip design. Handsomely bankrolled by the Abu Dhabi government--$10 billion so far, for a 68% stake--the new company has acquired Singapore's Chartered Semiconductor and also includes AMD's fab in Dresden, Germany. With these assembled fabs, GlobalFoundries is readying to face off with contract manufacturing giant TSMC in Taiwan.

The GlobalFoundries development represents the largest greenfield chip-manufacturing investment in the U.S. in years and offers at least a pause in the seemingly inexorable trend of manufacturing migrating overseas. Intel has also thrown down some chips: the company announced on Oct. 19 that it would spend $6 billion to $8 billion upgrading U.S. fabs and building a new R&D fab in Oregon.

The U.S. share of global chip-production capacity, once 100%, has been sliding for years: from 25% in 2005 to 14% in 2009. This hasn't seemed a problem because U.S. firms still retain a nearly 50% global market share (in sales), and roughly 70% of U.S.-designed chips are still made at home. This generated $48 billion last year, making the semiconductor industry a contender for the country's top industrial exporter, along with aircraft makers.

Much of that is due to Intel, which has managed to retain 75% of its production in the U.S., even though 75% of its revenue comes from overseas. Its new U.S. plant is specifically aimed at catching up in chips for smart mobile devices, a sector whose importance Intel has admitted it was slow to grasp.

But even this historical grounding in the U.S. is beginning to erode. In recent years, Intel has opened plants in Ireland and Israel, and its newest plant will open in Dalian, China, Oct. 26. "We're the last one standing in the United States," Intel CEO Paul Otellini told an audience at the Council on Foreign Relations (CFR) in October. "No one else has built a new factory in five or 10 years. Everyone is building it either offshore or through joint ventures somewhere else. And if this is the most important technology of the 21st century--semiconductors--and the first derivative is negative relative to building new factories here, it ain't good."

The industry now has its hand out, warning that unless the U.S. is willing to underwrite new fabs, the move overseas will accelerate. From 75% to 80% of new equipment is headed overseas, warns George Scalise, president of the Semiconductor Industry Association. Of the 27 fabs that closed last year, 15 were in the U.S.

The typical range of financial inducements, from outright grants to tax credits and holidays, trims the cost of building and operating a fab over 10 years by about $1 billion. In a job-short and export-challenged economy, it seems like a no-brainer. But in postbailout America, using federal funds for private industry is not getting a lot of traction.

Big mistake, says William Spencer, an industry veteran and vice chairman of a National Research Council board on science, technology and economic policy. He adds, "The U.S. government is out of step with the rest of the world in government-private partnerships to develop innovations for economic growth and job development, and we are jeopardizing our future in high technology."

Otellini has been making the case for leveling the playing field through public incentives, lower taxes (or tax holidays) and fast-track permitting. Intel's new Chinese plant, he points out, will be its lowest-cost manufacturing site. "It has nothing to do with labor," Otellini told the CFR audience. "It has to do with the capital grants, the equipment grants, the tax holidays that they gave us there."

New York's decision to sweeten GlobalFoundries' plans by $1.2 billion was a decisive factor in winning the investment. And New York's outlay goes beyond the plant. The University at Albany--SUNY's College of Nanoscale Science and Engineering (CNSE), financed with more than $900 million in state seed money, is putting a marker down. The college has attracted $6.5 billion more from major players, including IBM, Applied Materials, Sematech, AMD and GlobalFoundries, which time-share research facilities. The goal is to push beyond current technical boundaries. "We are the last line of defense for the U.S. semiconductor industry," argues Alain Kaloyeros, CNSE's CEO.

There are others in the U.S. industry who contend that this manufacturing migration overseas is a natural progression and not a problem as long as so-called headquarters skills--particularly chip design, where most of the intellectual property is thought to reside--remain in the U.S. Besides, argues Bolaji Ojo, business editor at Electronic Engineering Times, a leading industry publication, "this [migration] of the entire extended supply chain has been going on for 15 to 20 years, and it's not possible at this point to put that back together. It's gone."

For the increasingly fabless U.S. industry, the New York foundry substantially expands the open-to-all-customers capacity for chip designers. "It's also great from a risk-mitigation standpoint," says Norm Armour, vice president and general manager of the New York plant. "When a major earthquake hits Taiwan, guess what. It knocks about 20 wafer fabs off-line for days or weeks." For these reasons GlobalFoundries, which will rely mostly on AMD for the bulk of its initial production, should be able to add customers and keep its capacity utilization high.

Linley Gwennap, president of an independent industry consultancy, says that even when other countries can match U.S. technical skills, integrating them into a globally competitive company is another order of challenge. "We are probably 10 years away from seeing a Chinese or Indian semiconductor company that's really competing against the other leading companies in the world. But there is no question it is going to happen."

By then the U.S. firms expect they'll be on to the next big thing: nano. That means getting smaller, more powerful, with greater functionality, while decreasing power usage. The industry has done a remarkable job of extending Moore's law of doubling chip density and functionality every 18 months. Indeed, bleeding-edge technology is already mind-bogglingly small. Circuitry can be etched on a chip that is 28 nanometers wide (GlobalFoundries' starting point); 5,000 of these could fit into a single strand of human hair.

Getting beyond this, as the industry inevitably will, is an ever tougher and more expensive slog, especially as the physical limits of the current silicon-based technology are reached and the true nano era of single atoms begins. Kaloyeros points out that no rocket could get off the launchpad today without chips. "This is really rocket science squared." It's very likely that CNSE will design such nano chips; what's not so certain is who will make them and where.

Read the full article:,9171,2028089,00.html