Will Intel's Planned China Fab Start An Offshore Exodus?

The U.S. chip industry has just taken one on the chin, said iSuppli analyst Len Jelinek.

Mike Clendenin, Contributor

March 28, 2007

8 Min Read
InformationWeek logo in a gray background | InformationWeek

Champagne sat chilling while guests milled about the Great Hall of the People in Beijing awaiting the honored guest, Intel Corp. chief executive Paul Otellini. Everyone knew what he'd say--that Intel would build its first Asian fab in China. But there was still an air of latent disbelief.

Five years ago, the only time Intel would have considered 12-inch in China would be at a pizza parlor near its Beijing headquarters. Now the semiconductor giant is looking to spend $2.5 billion to build a plant using the supersize wafers, albeit with lagging 90-nanometer process technology. Still, as the bubbly passed round, one couldn't help but wonder how many more IDMs would follow Intel into China, and what that might mean to a U.S. and European chip-making industry on the decline?

"The U.S. chip industry has just taken one on the chin," said Len Jelinek, an analyst at market research house iSuppli Corp. (El Segundo, Calif.).

Otellini made his view as clear as the crystal he drank out of: "It is no secret that China is at the forefront of a remarkable surge in both market growth and innovation," he said. "Today's announcement sends a message that the Chinese market is very important to Intel."

Indeed. As a consumer, China is already Asia's top PC and handset market, and Intel's second-largest market. Consumer electronics gadgets are also a big draw among the growing middle class. In production, China is busting traditional emerging-market development models. It combines a seemingly inexhaustible supply of cheap labor for low-end manufacturing with a lot of supersmart, hard-working people eager to play at the high end. Add a government bent on becoming a leader in innovation, and willing to offer sweet deals to draw industry leaders, and it's an attractive mix for Intel.

Through action and word, Intel is spelling out America's future place in chip making. The United States can no longer assume that its role as a technology leader will translate into big plants. "In general, IC manufacturing has been drifting slowly away from U.S.- and Europe-based operations. We expect the evolution to continue to occur," said Bill McClean of market research house IC Insights Inc. (Scottsdale, Ariz.). Intel's Otellini pointed out that of the last three fabs Intel has built or announced, two have been outside the United States, in Israel and now China.

To put a finer point on how Intel sees China, consider this: Intel does not build trailing-edge fabs. It has not, in 15 years, built at a greenfield site. It is not enamored with China's poor record on intellectual-property protection. Yet it chose China anyway. And while the company will spin up the fab by making chip sets, Otellini hinted that eventually CPUs--Intel's crown jewels--will roll off the lines, too.

During the next five years, industry observers say more IDMs and foreign foundries will build fabs in China. Analyst McClean believes the strength of the IC manufacturing base in China "depends on foreign fabs," and he sees local operations, like Semiconductor Manufacturing International Corp. (SMIC) and Hua Hong NEC, playing a smaller role.

Things won't go gangbusters, though. By 2011, China will crank out just 3 percent of global chip production, up from 1.5 percent in 2006, according to IC Insights. Over the long term, however, China's attraction as a fab hub should steadily improve. If things go well in Dalian, the city in northeast China where Intel will put Fab 68, then it will almost certainly build more fabs there to attain economy of scale.

With an expanding 12-inch wafer fab run jointly by Hynix Semiconductor and STMicroelectronics, and now Intel's play, there is cause for renewed optimism among equipment sellers with a strong presence in China. Spansion is also on the hunt for a fab in China, according to a source at that company, and it may also make sense for the Intel-Micron flash joint venture to consider a plant here.

Spending by existing fabs will remain flat to moderate in the next two years, said Samuel Ni, senior China market analyst at Semiconductor Equipment and Materials International. "But with Intel's tug, we expect that total capex and fab equipment spending in 2008 to 2009 may reach the next peak in China," he said. The group will publish an update on China capex in the coming weeks.

Intel expects Fab 68 to be its lowest-cost site by the time it comes online. For the first time, the company won't ramp a fab with its "copy exact" strategy--using the same tools and process recipes to speedily get a facility up and running. That's been a bedrock principle of Intel manufacturing.

Instead, Otellini said the company will experiment with "new technology" to reduce costs. He declined to elaborate. Some U.S. engineers will come in to oversee the facility, but it won't be on the scale seen at Hynix-ST Semiconductor in Wuxi, China. That facility is one of the fastest ST ever brought online, but it happened largely because partner Hynix shipped in about 500 engineers from South Korea.

"By 2010, the 90-nm and even the 65-nm processes will most likely see continued improvement--i.e., learning curve--productivity enhancements, etc. So if Intel is focused on cost, there are probably better ways to get there than the original process developed almost four to six years prior," said Joanne Itow, an analyst at Semico Research Corp.

With the blessing of the world's largest chip maker, anything seems possible for China's chip future. Yet not all is probable. A choke point in China's grand design is getting its hands on advanced technology, especially from U.S. manufacturers. Even Intel can't surmount this hurdle.

To get a deal in place, Intel had to promise the U.S. government to use technology that is a few generations behind. The United States has granted Intel a one-off approval to transfer its 90-nm process technology to China and bring in fab equipment from U.S. vendors such as Applied Materials Inc. and Varian Inc. Plans call for the fab to use 90-nm technology, but by 2010 it will likely be 65 nm. Yet by that point, Intel will be producing elsewhere at 32 nm, and testing at 22 nm.

So an Intel fab in China, although a coup for the Chinese, does not signal an overall loosening of U.S. export policy on technology.

In general, there is no standardized policy for making decisions about whether to grant export licenses for semiconductor-related sales to China. Each application is reviewed by an interagency group that includes Commerce and Defense Department officials, and the criteria usually include the reliability of the end user, the stated end use, the level of technology and the export compliance record of the exporter. In all those areas, Intel scores well.

Despite its OK on the Intel deal, the U.S. government is moving to tighten export policy on technology. The changes, championed by the Bush administration, were scheduled to fall into place as early as this month and are intended to cast a wider net over commercial technology products that might make a "material contribution" to the advancement of China's military.

The technology industry has blasted the plan and deluged the administration with more than 1,000 pages of feedback during a six-month public comment period that ended in December. The administration, however, may be taking a second look at the controversial plan (see story, page 28).

In other parts of Asia, Intel's move is raising eyebrows. Foundry giant Taiwan Semiconductor Manufacturing Co. has already said the Taiwan government should move more quickly to loosen restrictions in light of the deal. India's chance for an Intel fab was always a long shot, but there is still hope AMD may come through there in the next few years. Only fanciful thinkers believe Japan will locate an advanced fab in China.

Because Intel lives and dies on advanced technology, it will no doubt continue to build more advanced fabs at home. But the United States will need to think hard about the kind of teaser incentives required to compete against other governments--and not only in Asia. Across Europe and in the Middle East, governments are ready to pony up hundreds of millions for fabs as a way to kick-start high-value growth. Some U.S. states are fighting back--for instance, New York promised $1.2 billion in incentives and associated investments to lure AMD's fab. Yet Otellini said that it costs Intel $1 billion more to build in the United States than in countries offering tax breaks and incentives.

McClean, at IC Insights, believes that figure is exaggerated. "The U.S. is not totally inept in this area. Given that about 75 percent to 80 percent of the cost of a fab is due to equipment costs, which are essentially the same price worldwide, a billion-dollar differential seems too high," he said.

With the new fab, Intel's investment in China will total $3.8 billion, making it one of the country's largest foreign investors.

 See related image

Never Miss a Beat: Get a snapshot of the issues affecting the IT industry straight to your inbox.

You May Also Like


More Insights