close

OP-ED: Computer chips require global partnership

By Gary Stout 5 min read
article image -
Gary Stout

Several years ago, we were traveling on an Egyptian tour bus listening to our guide as we traveled from Cairo to Alexandria. He was explaining the sorry state of the Egyptian economy and made what I thought was a joke that “Egypt’s only real export is sand.”

It turns out that our guide was actually informing us of the first step in the long, worldwide process in producing computer chips. Since our trip to Egypt, I have learned more about the incredible collective international achievement that has become the semiconductor/chip supply chain. With artificial intelligence and the firing of Jensen Huang at Nvidia dominating the news, some basic information on computer chips may be helpful.

Before we begin our journey down the computer chip assembly process, some definitions are in order. Semiconductors are most often made from silicon (sand) and can conduct electricity. Wafers are thin slices of semiconductor material used to build chips. Computer (micro) chips are small devices that hold integrated circuits. An integrated circuit gives the chip functionality.

Many microchips begin in the white sand fields of the Sinai Desert in Egypt. The sand is of an exceptionally high quality, 98.8% silicon dioxide. Large quantities are exported to Turkey, Japan, and China to be fed into large industrial smelting operations. At temperatures exceeding 3,452 degrees Fahrenheit, metallurgical grade silicon metal is produced.

To be usable in computer chips, metallurgical silicon must be transformed via chemical processes into electronic-grade polysilicon (99.9999999% pure). China manufactures a dominant share of global polysilicon. Only about 2.4% of the world’s total polysilicon is refined to this hyper-pure standard to build computer chips. The rest goes into Chinese-made solar panels.

The next process called “crystal growth” involves melting down the silicon. As the liquid cools around the crystal seed, it copies its perfect geometric lattice, creating a massive, single-crystal cylinder called an ingot. Major silicon wafer manufacturers like SUMCO (Japan), Global Wafers (Taiwan), and Siltronic (Germany) perform this function.

At these facilities the heavy, solid ingots are ground down to exact diameters (typically 300mm or 12 inches for advanced chips) and then fed into precision diamond wire saws. The wire saws slice the ingot into ultra-thin discs — silicon wafers -measuring less than a millimeter thick. These raw discs undergo rigorous chemical etching, lapping, and atomic-level mirror polishing to eliminate any surface micro-damage.

The silicon wafers are now ready to ship directly to chip fabrication centers (foundries). Taiwan Semiconductor Manufacturing Company (TSMC) is the largest chip foundry in the world with a market share of 73%. More importantly, it manufactures over 90% of the world’s most advanced chips (used for AI accelerators, smartphones, and data centers). The world’s most important end product manufacturers like Apple, Nvidia, AMD, and Qualcomm do not have their own foundries.

Foreign policy experts around the world realize that Taiwan’s role in producing most of the high-end chips is a major “choke-point” in the production process. If China decides to blockade or invade Taiwan, chip production would grind to a halt. Recently, the U.S. began efforts to “onshore” its own chip production by providing seed money to the American company Intel to build a foundry.

The chips are still not ready to be placed into a phone or server. The circular silicon wafers must be sliced into individual dies, tested for defects, and protected in a protective casing (packaged). While TSMC handles some advanced packaging in-house, a significant portion of wafers are shipped to other providers in Asia.

The production of the most advanced chips would not be possible without a new technology, perfected in 2018. The Dutch town of Veldhoven, near the border with Belgium, is home to the only factory capable of assembling a revolutionary machine that’s relied upon by the world’s biggest chipmakers. Advanced Semiconductor Materials Lithography (ASML) has a monopoly on the fabrication of EUV lithography machines. According to Chris Miller, assistant professor at the Fletcher School at Tufts University, “The machines that they produce are among the most complicated devices ever made.” Priced around $400 million, these ASML systems use larger, hyper-precise mirrors and faster stage accelerations to print smaller features and significantly increase transistor density for advanced AI chips.

We now return to Taiwan and the bottleneck this country creates in chip production. The previously mentioned TSMC operation, the world’s largest contract chipmaker, uses ASML’s systems to manufacture chips for companies like Apple and Nvidia.

Most end-product companies design the architecture and blueprints of their unique merchandize in-house. However, they outsource the actual physical fabrication to third-party contractors. The end-product is then shipped directly to hyperscalers or data centers like Amazon, Microsoft, and Google. Other finished devices go to the tech consumer giants like Apple, Samsung, and HP to be sold as smartphones, gaming consoles, and laptops.

To better understand the above summary, I highly recommend a recent New Yorker article with a very long title. “Computer Chips Are the Most Magnificent Things Humanity Has Ever Made” appears in the Oct. 26 issue. The essay traces the long journey of computer chips, gives the history of the chip, and provides many anecdotes from the key players.

A major takeaway for me was that any American attempt to totally make chips at home rather than overseas is illusionary. A chip in production can cross international borders more than 70 times before it’s shipped to a customer.

Gary Stout is a Washington attorney.

Starting at /week.