IBM researchers have announced a breakthrough in microchip manufacturing. Currently, the smallest circuits in commercial production have etchings of about 65 nanometers in width. But IBM says it has figured out how to produce integrated circuits at just 30 nanometers. According to the Los Angeles Times, IBM’s new technique could save the industry loads of money. Why are smaller microchips cheaper to make than bigger ones?
You can make more of them at once. To make a microchip, you first have to produce a very pure, sausage-looking hunk of silicon called an “ingot” that’s up to about a foot in diameter. Then you slice the ingot into very thin wafers, sprinkle an insulating material onto the surface of a wafer, and etch out a precise pattern using ultraviolet light. (You also apply conductive metals and other materials in multiple layers to build up the chip.) These patterns get repeated over the surface of the wafer as many times as they can fit.
When the etching and layering are done, you’ve got a silicon wafer that contains dozens or hundreds of microchips, kind of like a sheet of uncut baseball cards. (The chips get separated with a diamond saw or a similar piece of equipment.) The smaller the pattern, the more microchips you can cram onto a single wafer. At the same time, the cost of processing a wafer doesn’t really increase. Chip manufacturers use more or less the same amount of material and do the same amount of etching whether they’re drawing 500 tiny chips on a wafer or 100 bigger ones.
Another consideration is that silicon is never 100 percent pure. That means you’re guaranteed to etch some of your microchips on flawed material. Any defective microchips will have to be tossed out. If you know you’re going to get a few defects per wafer, you’d rather lose a few small chips—and throw out a small amount of silicon—than a few big ones.
Even if a manufacturer does have to upgrade some equipment to etch smaller chips, it will likely benefit from the increase in production volume. A manufacturing facility—called a “fab”—costs billions of dollars to set up. Since the cost of producing a chip reflects the cost of building a fab, any increase in capacity down the line will likely lower the overall cost per chip. IBM’s discovery will save money because it improves on existing technology without changing it entirely. Manufacturers wouldn’t be as keen on a brand-new method that requires a fully revamped fab.
Got a question about today’s news? Ask the Explainer.
Explainer thanks Kevin Krewell of the Microprocessor Report.