After decades of decline, a new boost in US manufacturing is garnering considerable media attention. In particular, the semiconductor industry has been prioritized for US government support through the CHIPS Act initiative with some US$50 billion worth of state funding.
But what will be needed to increase and sustain US high-technology manufacturing? A serious resurgence of advanced manufacturing (chips being the most demanding) will require much more than investing in more sophisticated equipment in new plants.
It will require training a new generation of highly skilled personnel to operate such plants successfully. While increasingly sophisticated technology is key to much of competitive manufacturing, it is productive only with staff with very specialized training to operate in complex plant environments. Badly managed mechanization will hinder rather than promote value creation.
I learned from experience in semiconductor manufacturing. Early in my career at RCA, I was tasked with designing and operating one of the first silicon transistor factories to manufacture the 2N2102, a transistor I had developed for use in building computer and other electronic systems.
I outfitted the factory with equipment scaled from my laboratory. At the time, no commercial production equipment existed. For example, the optical lithographic equipment was built by the local photography shop in Somerville, New Jersey, for a few hundred dollars.
When production volume needed to increase, new equipment was required and we purchased new, more automated commercial production equipment then coming to market. I had hoped that such equipment would increase production volume and yields.
However, volume increased but product yields declined. The cause, we discovered, was inadequate process definition for the automated equipment.
We found that with the original manual processes, the production technicians introduced changes as they deemed necessary to maintain quality and production rates in the face of small changes, for example, in the temperature of chemical solutions.
The new machines were designed to operate under fixed pre-set conditions and changes were not automatically adjustable. To make the factory operate, we had to invest time and effort in defining all of the production elements and avoid human intervention randomly introduced to correct anomalies.
The machines had to be programmed with great precision for desired results. This required much process research to understand variables.
Here, I learned a valuable lesson in production management—the importance of very precise process definition and control. As new, more automated equipment was introduced, product yields frequently declined initially.
They did not improve until the production process was refined to new levels because new automated equipment required new levels of process knowledge and control.
To enable successful production required a close working relationship between production process engineers and equipment operators. These engineers had to be trained to fully understand the technology. This was a new engineering discipline.
This necessitated change in production methodology was costly and hard to accept by production managers trained in the old days of low levels of automation. But change they did as a new generation of production managers entered chip manufacturing.
However, I found that, given the same equipment, plant performance was a strong function of local management quality and staff training. The variables to be controlled and adjusted with automation were practically endless and the plants had to develop their own process engineering skills to perform economically.
In effect, starting with devices designed in the laboratory, moving them into production was a whole new endeavor requiring talent as valuable as that of the original design team.
It is evident that as chips increased in complexity, production became extraordinarily demanding in capital and human resources. For my original transistor, the device dimensions were in fractions of a centimeter and a photoshop could make the equipment to form the dimensions.
Today, integrated circuit chips have billions of transistors interconnected with dimensions approaching atomic ones. Whatever challenges I faced, today’s plants have them in far greater complexity and cost.
Fast-forward to 2024, and my simple homemade lithographic tool has evolved for the most sophisticated chip production to a huge piece of ultraviolet laser-powered equipment produced by ASML (the sole producer globally), which sells for about $300 million and requires specially trained staff to operate.
Faced with such costs and management challenges, it is easy to understand why the leaders of so many chip companies decided that manufacturing was too challenging and outsourced production to Taiwan’s TSMC, which is exclusively committed to chip manufacturing.
TSMC’s success is not based on the invention of unique equipment. Rather, it is rooted in outstanding management of human and capital resources. The company trains its staff to a high level of performance and operates its plants to get the best possible performance from its costly equipment.
A new fully equipped plant costs $20 billion but it is the highly trained management and staff that make it work. At this time, TSMC produces over 90% of the highest-performance chips in the world. Anyone looking to compete has to invest in the human resources needed, not just the equipment.
Henry Kressel is a technologist, inventor, author and industry executive. He is a long-term private equity investor in technology businesses. Incidentally, the original 2N2102 product is still commercially available.