Practical Chip Design

Semiconductor Manufacturing International Corp. (SMIC, to most of us) announced its financial results today, and the press quickly reported softening revenues, sharply lower guidance for the fourth quarter, and a retreat on capital expenditure plans. But a couple of less bottom-line points caught our attention.

One is that SMIC is continuing to follow the big guys—at a discrete distance, of course—in process migration. The statement released with the figures said that the company has had more than 20 product tape-outs at 65 nm, all in various stages of qualification. As has been reported before, the company is collaborating with IBM on 45 nm, and has recently received the export licenses it needs for 32 nm process development.

Even more interesting is the rapid shift in the geographic make-up of SMIC's business. The foundry has been, to a great extent, a supplier of low-cost, unambitious wafers to IDMs and fabless companies outside China—a thrift shop, if you will. But the percentage of business from fabless companies inside China is growing rapidly, even after you take into account that SMIC is disposing of its DRAM business as quickly as possible, raising the percentages for all its logic business.

The company said that shipment quantity—by which I presume they mean unit volume—to Chinese customers increased 28 percent between the second and third quarters. Asia-Pacific, excluding Japan, has grown from roughly a quarter of revenue in the third quarter last year to slightly over a third of revenue in the current quarter. The company goes on to cite some specific projects that are either under way or shipping: a decoder chip for the Chinese HDTV standard, a chip for the China Multimedia Mobile Broadcasting standard, and a TD-SCDMA chip for the country's next-generation mobile phone standard. None of these is an unambitious project, and all are demanding on IP and on foundry quality.

Two things seem clear from this snapshot of the biggest Chinese foundry. First, the company has no intention of falling off the treadmill of process development, even if it has little prospect of catching up right away. Second, even though demand may be softening among Chinese chip users as it is everywhere else, the ability of Chinese fabless companies to produce sophisticated designs, and to take those designs to market successfully, is growing. Today, the most sophisticated chips are being designed under the shelter of proprietary Chinese communications and consumer-broadcast standards. But these will be learning experiences that could lead to a tier of fabless companies in the People's Republic able to take the world on head-to-head on complex consumer and communications SoCs. Watch this space.