Chipmakers like TSMC and Intel have invested billions of dollars into factories, but finding qualified staff is a challenge. Image: Handout

The semiconductor industry was invented and developed in the United States, yet most of the highest-performance chips are now produced by Taiwan Semiconductor Manufacturing Company (TSMC), with more than US$70 billion of annual revenues solely from chip manufacturing of devices designed by others.

Such devices are key enablers of all high-performance electronic systems. Faced with defense-system concerns, the US government is providing $50 billion to fund the most advanced new chip production plants domestically. The idea is to revive high-performance chip production in the US.

Based on my experience, I believe that high-end chip manufacturing is the most challenging in the world. It costs billions of dollars to build such plants, but that’s only the beginning: The qualified staff to operate them cannot be purchased – they need to be trained and developed in a production environment.

Such a program will not succeed without a major increase in the number of sophisticated technologists capable of operating such plants.

TSMC’s much-heralded American plants, to be built with US government subsidies, have been delayed repeatedly – something that doesn’t happen in Taiwan. A plant scheduled to open in 2026 won’t be operational until 2027 or 2028, TSMC said on January 19. The opening of another plant was delayed to 2025 from a planned 2024 opening. The company blamed a shortage of skilled workers.

I have been involved in the design and operation of chip plants since the 1960s. They require uniquely qualified people trained over periods of not months but years. The equipment to build such factories is commercially available and is extremely complex and automated.

But the key manufacturing enabler is the people who operate it. It takes years to develop such expertise, and it is rare in the United States, as such manufacturing has largely gone overseas, leaving Intel as the only domestic manufacturer of high-end chips.

The United States still leads in the design of the highest-performance chips, and the recent success of Nvidia with leadership in AI-enabling chips (produced by TSMC) is an indication.

With the exception of Intel, the US industry was happy to outsource the production of the most sophisticated integrated circuits to TSMC, which proved to be a reliable leading-edge production company.

Gradually, TSMC acquired the facilities and skill to produce the best leading-edge devices designed by companies globally. It is today a very profitable high-volume chip manufacturing technology leader. With scale comes market position and economics of production, and the company is very profitable. Any potential competitor has many barriers to overcome, and the biggest ones are in management and staff expertise.

Manufacturers’ challenges

What makes high-end chip manufacturing so challenging are two factors. The first is the shrinking dimension of transistor core devices that are now in the low-nanometer range (approaching interatomic dimensions), and second, the need to manage a production process where different very thin chemical materials are deposited on a 12-inch-diameter (305-millimeter) silicon base.

Typically, more than 100 such films must be deposited and patterned using a variety of techniques, and the specific configurations must be controlled for different device structures. Within these three-dimensional structures are patterns that need to be created for circuit interconnections.

So you have a production process that requires both thin-material technologies and accurate patterning of near-atomic dimensions. A finished chip may have several billion transistors interconnected per square inch of surface.

Controlling such a production process is arguably the most difficult manufacturing problem in the world. Different products require different production steps that number in the thousands from start to finish. There can be no errors in the production process, as errors will impact the performance of the devices.

You can imagine the difficulty of entering such a challenging production business. It is no wonder that chip design companies have chosen to avoid owning such production plants when TSMC does such a great job economically.

And now comes the US government offering financial support for entrants into this production business. Only the hardiest will make the commitment, but staffing will be the biggest challenge.

Operating a chip plant with this complexity requires people with a high level of technology skills. Above all, it requires people well-schooled in the chemical, mechanical, and electronic technologies underlying the production process.

In a company like TSMC, it is production technologists that are the ultimate decision-makers because they best understand the tradeoff between process yield and device performance. It is that effective decision process that is reflected in the company’s production performance and cost.

Is there a way to meet the US government’s goal? Only if a serious investment in the right educational programs is made, possibly through state-funded community-college training with rigorous standards.

This is not a short-term endeavor. I believe that training for the most modern manufacturing industries will, in general, benefit the US economy and reverse what has been a continuing decline in domestic high-technology industries.

The combination of capital and a highly trained workforce is needed. Without much more of the right vocational education, US taxpayers’ money will be wasted.

Henry Kressel is an inventor, technologist and author, as well as a long-term private equity investor in technology companies.

Join the Conversation

1 Comment

  1. The clique of rent seekers that control the US has no real interest in rebuilding the production-based economy upon which all wealth is created. If it were serious, a 21st Century National Defense Education Act would be rushed through Congress. The effectively idle youth, and not so young, would be dragooned into scientific, technological, and vocational pursuits. Eventually, the necessary human capital would be created, but it can’t happen overnight.

    So do we see this happening? No, absolutely not. The US is where it is precisely because rent-seekers have no interest in production, only in collecting rents. It is like Britain living off extraction of the wealth of India in the latter decades of the 19th Century while nervously eyeing the rising Bismarkian Germany across the Channel, while thinking of war.

    So what is going on? China, of course. China creates wealth; the US consumes wealth without offering any in exchange. This gravy train is about to come off the rails for the reasons we all know.

    So what to do? Destroy China, of course. This would ideally mean a low-cost, war, a la Ukraine, against China in the Taiwan Strait and/or the South China Sea, and a blockade of the Straits of Malacca. And crucially, fabs of TSMC would be destroyed to keep them out of the hands of the PRC.
    This would be bad for everyone. However, if the US effectively duplicated TSMC’s fabs in the US, it could be a win, owning the sole source of sub 5nm chips! It would be like how the destruction of the Nord Stream pipelines created a bonanza for US LNG suppliers.
    So that is what the rush to build advanced semiconductor manufacturing in the US is all about. Not rebuilding a wealth creating economy, but preparations for war.