Over the past few weeks, headlines have been screaming about the imminent threat and potential impact of U.S. import tariffs on semiconductors. To be honest, I don’t think the implementation of these tariffs will happen because they will cause such a significant supply chain disruption, and its annoying impact is still too new in the memory of Covid-19. Who can forget the thousands of unfinished cars trapped in the automaker’s lot. Of course, no one wants to repeat it!
That said, I believe that becoming more resilient and self-reliant in the semiconductor manufacturing sector still keeps American businesses and the U.S. economy in general, and I appreciate these efforts. Here we will look at why this self-reliance is so important, especially in the United States’ ability to maintain (currently narrow) leadership in state-of-the-art artificial intelligence (AI).
The core of AI competition is Chips Race
Semiconductors are crucial to powering servers that train AI models, because training these models requires a specialized intensity that only semiconductors (with traditional processors) can provide. It is estimated that by the end of this year, AI-related semiconductors will account for 19% of the global semiconductor market, a significant increase from the number of people held in 2017.
The increased dependence on semiconductors for AI means that the less the United States relies on foreign entities for semiconductor supply. With the global AI race heating up, domestic semiconductor production has brought significant benefits such as solid economic and national security and technological independence. Currently, there is a bill passing Congress called the Semiconductor Supply Chain Act 2025, which has bipartisan support and aims to reduce reliance on unpredictable foreign supply chains.
What should we do?
In response to the threat of possible U.S. import tariffs, many expressed concern that in its current state, the U.S. has enough capacity to handle soaring semiconductor demand driven by the buildings generated by AI and AI data centers. Business uses of AI, such as coding and software development, are particularly at risk. Any disruption in semiconductor access can cause a chain effect across applications that depend on applications including autonomous vehicles, edge computing and robotics.
The U.S.’s ability to drive innovation in semiconductor-dependent industries, including AI, will require accelerated material discovery. The “old ways” of material discovery and adoption are often concentrated in overseas foundries and involve multi-step processes such as lithography, etching, deposition and cleaning rooms. This can be a slow and expensive process that leads to design cycles and a lot of waste of materials.
To better meet domestic semiconductor needs, the United States must take advantage of advances in chip design, one technology is direct local atomic layer processing. This is a digital, atomically accurate manufacturing process that builds devices directly from atoms, eliminating the need for many steps involved in traditional manufacturing processes while reducing complexity and waste. It provides unprecedented flexibility and precision for design and prototypes, including AI semiconductors.
By enabling precision of atomic scale and control over material processing, technologies such as direct local atomic layer processing can significantly accelerate design cycles and prototyping, helping to find new materials or combinations of materials that can meet AI’s growing computing needs.
Increase domestic manufacturing industry while working on the environment and human health
As an additional (not irrelevant) benefit, new technologies can also significantly reduce the environmental impact of semiconductor manufacturing. So far, the industry faces serious dilemma due to its oversized environmental footprint, which contributes greatly to greenhouse gas emissions, water consumption and chemical waste, especially the toxic “forever chemicals” known as PFA. These chemicals contaminate water, don’t break down and stay in the environment (and people!) for decades.
No wonder recent federal actions such as the Buildings Act in the United States Act and the Bargaining Chips Act have caused significant environmental problems. By cutting the time required to design, prototype and manufacture chips and eliminating the need for chemically intensive clean room environments, new technologies can be the answer to meeting and using home resources to responsibly expanding their needs without damaging the environment and human health.
Utilize collective resources in the United States
In addition to deploying new manufacturing technologies, the United States must update its holistic approach. This means moving from a model of massive offshore production to a handful of billion-dollar foundries to leverage the nation’s comprehensive and abundant arsenal to collaborate, accelerating discovery and supporting the entire “lab-to-fab” process (research, prototype and manufacturing). This can all be achieved while keeping the cost at checking and integrating enabled technology directly into the infrastructure of these organizations.
Looking to the future
The relationship between AI and semiconductors is indeed symbiotic. As we mentioned, semiconductors are essential to power servers that train AI models. On the other hand, AI is significantly accelerating the discovery of semiconductor materials by leveraging machine learning to predict the properties of new materials and accelerate the design process. This approach is called inverse material design, allowing researchers to design materials with specific target properties, such as improving conductivity, energy efficiency, and sustainability.
Accelerating the discovery of new materials remains one of the toughest challenges in semiconductor manufacturing, although demands on AI semiconductors are particularly demanding as the industry seeks to continuously improve computing power, efficiency and speed while reducing chip size.
Although AI can be used to predict the properties of new theoretical materials, these breakthroughs are traditionally limited by the slow pace of physical verification. New technologies can be used to support high-throughput experiments, helping to close the gap. Achieve faster, more targeted material development and ultimately unlock next-generation materials. Combining new technologies such as direct atomic layer processing with the power of artificial intelligence has the ability to make magic, drastically accelerate the development of breakthroughs that have never been considered possible, all concentrated within the U.S.’s own national borders.
