From invention to industry
The invention of the first working chips in the late 1950s answered the question of whether a whole circuit could live on one piece of semiconductor. The next question was how to make such devices reliably, cheaply and in quantity. Most of the history of the microchip since then is really a history of manufacturing: better materials, cleaner factories, more precise tools and ever finer detail.
This article follows that production story. The question of who invented the integrated circuit, and how it led to the microprocessor, is a separate topic.
Making a chip, in outline
Although the details are highly technical, the basic process has stayed recognizable for decades.
- A very pure crystal of silicon is grown and sliced into thin round wafers.
- The wafer’s surface is coated with a light-sensitive material.
- Light shines through a patterned template, called a mask, to expose the design. This step is known as photolithography.
- Chemical treatments etch away or alter exposed areas, and other materials are added.
- The process repeats for many layers, building up transistors and the wiring that connects them.
- The wafer is cut into individual chips, which are tested and packaged.
Each wafer carries many copies of the same design, so improving the process helps produce more working chips at once.
The planar process and early growth
A key early advance was the planar process, developed at Fairchild in the late 1950s, which built components on a flat surface and protected them with a layer of oxide. It made it practical to put many parts on one slice of silicon and to connect them in a repeatable way. This approach underlies most chip making ever since.
Early chips in the 1960s contained only a small number of components, a level later described as small-scale integration. Over time engineers used terms such as medium-scale, large-scale and very-large-scale integration to describe growth in the number of components per chip. The labels are informal, but they give a sense of how quickly density rose through the 1960s, 1970s and 1980s.
Why chips kept getting denser
In 1965, Gordon Moore, then at Fairchild and later a co-founder of Intel, published an observation about how the number of components on a chip was increasing year after year. It was later adapted into a rule of thumb that is often phrased as transistor doubling at regular intervals. It was never a law of nature. It worked more like a target that the industry organized itself around, and the intervals quoted have varied over time.
Several developments made the rising density possible.
- Better lithography. Improvements in light sources, lenses and masks allowed smaller patterns to be printed. In recent years the industry has adopted extreme ultraviolet light for the finest layers, a technically demanding step.
- Cleaner factories. A single dust particle can ruin a tiny circuit, so chip factories use cleanrooms with tightly controlled air. Standards rose steadily as features shrank.
- Better materials. Refinements in silicon purity, insulating layers and metal wiring helped. Copper wiring and new insulating materials were adopted in various generations, and manufacturers also changed transistor shapes to keep them working as they shrank.
- Larger wafers. Moving to bigger wafers over time let factories produce more chips per batch, lowering cost per chip.
Process names and what they mean
Manufacturers describe generations of technology with a “process node” name, historically given in micrometers and then nanometers. For many years this number reflected a real physical dimension. In more recent generations, the name has become more of a marketing and naming convention, and numbers from different companies are not directly comparable. A careful reader should treat node names as labels rather than precise measurements.
A global and costly industry
As processes became more complex, chip making consolidated. Building a modern factory costs a very large amount, and only a small number of companies can afford leading-edge production. A common pattern is the split between companies that design chips and companies that manufacture them under contract, which developed from the late 20th century onward. Equipment suppliers are also specialized, and a few firms provide tools that are essential for leading production.
This concentration has made supply chains a matter of public concern. Disruptions at a few sites can affect many products, from cars to consumer devices. That topic belongs to economics and policy, but it shows how far chips have moved from laboratory curiosities to infrastructure.
Beyond simple shrinking
In recent years, simply making transistors smaller has become harder and more costly. Engineers have turned to other strategies, including three-dimensional stacking, combining several smaller chips in one package, and specialized designs for tasks such as graphics and machine learning. Memory chips and processor chips have both followed these trends in different ways, and technologies like flash memory depend on stacking many layers.
Whether the historical rate of progress can continue is debated. Some observers expect gradual slowing, and others point to new techniques. It is safest to say that the future pace is uncertain.
The takeaway
The microchip did not improve through a single breakthrough but through thousands of manufacturing refinements, many of them invisible to the public. That patient, cumulative engineering is a big part of why a device in a pocket can contain circuitry that earlier generations would have found hard to imagine.