For the past several weeks, headlines have been screaming concerning the looming threat and potential impact of U.S. import tariffs being imposed on semiconductors. Truthfully, I don’t think implementation of those tariffs will ever occur because they might lead to such significant supply chain disruption, the nasty effects of that are still all too fresh in our memories from COVID-19. Who can forget the tens of hundreds of unfinished cars left stranded in automotive manufacturers’ lots. Definitely, nobody wants a repeat of that!
That said, I consider it still behooves U.S. businesses and the U.S. economy overall to turn into more resilient and self-reliant in the realm of semiconductor manufacturing, and I applaud these efforts. Here, we’ll examine why this self-reliance is so necessary, particularly when it comes to the U.S.’s ability to keep up its (currently narrow) leadership in state-of-the art artificial intelligence (AI).
The AI Race Is, At its Core, A Chips Race
Semiconductors are crucial for powering the servers that train AI models, as training these models requires a specialized strength that only semiconductors (versus traditional processors) can deliver. It’s estimated that by the tip of this 12 months, AI-related semiconductors will account for 19 percent of the whole semiconductor market worldwide, a big increase from the seven percent held in 2017.
Increased reliance on semiconductors for AI means the less the U.S. relies on foreign entities for semiconductor supply, the higher. As the worldwide AI race heats up, domestic semiconductor production offers significant advantages like bolstered economic and national security, in addition to technological independence. Currently, there’s a bill passing through Congress called the “Securing Semiconductor Supply Chains Act of 2025,” which has bipartisan support and is aimed squarely at reducing reliance on unpredictable foreign supply chains.
How Do We Do It?
In response to the specter of possible U.S. import tariffs, many have voiced concerns that in its current state, the U.S. is ill-equipped to handle the skyrocketing semiconductor demand being driven by generative AI and AI datacenter build-outs. Business uses of AI, similar to coding and software development, are especially in danger. Any disruption in semiconductor access could induce a ripple effect across dependent application areas, including AI and downstream markets like autonomous vehicles, edge computing and robotics.
The U.S.’s ability to drive innovation across semiconductor-dependent industries, including AI, would require an acceleration of materials discovery. The “old way” of materials discovery and adoption was typically concentrated in overseas foundries and involved multi-step processes like photolithography, etching, deposition and clean rooms. This could be a slow and expensive process, resulting in lengthy design cycles and significant materials waste.
To higher meet semiconductor demand domestically, the U.S. must reap the benefits of advancements in chip design, one technique being direct local atomic layer processing. This can be a digital, atomically precise manufacturing process that builds devices directly from atoms, eliminating the necessity for the various steps involved in the normal manufacturing process, while reducing complexity and waste. It offers unprecedented flexibility and precision for designing and prototyping a wide range of microdevices, including AI semiconductors.
By enabling atomic-scale precision and control over materials processing, technologies like direct local atomic layer processing can significantly speed up design cycles and prototyping, helping to seek out recent materials or combos of materials that may satisfy the ever-growing compute needs of AI.
Increasing Domestic Manufacturing While Remaining Committed to Environmental and Human Health
As an extra (and never inconsequential) profit, recent techniques can even dramatically lessen the environmental impact of semiconductor manufacturing. Up to now, this industry has faced a serious dilemma as a result of its outsized environmental footprint, contributing significantly to greenhouse gas emissions, water consumption and chemical waste, particularly toxic ‘perpetually chemicals’ often known as PFAS. These are chemicals which pollute water, don’t break down and remain within the environment (and in people!) for many years.
It’s no wonder that recent federal actions just like the Constructing Chips in America Act and the CHIPS Act have raised significant environmental concerns. By slashing the time needed to design, prototype and manufacture chips – and eliminating the necessity for chemical-intensive clean-room environments – recent techniques will be the reply to satisfying demand and scaling responsibly using domestic resources, and without compromising environmental and human health.
Harnessing the U.S.’s Collective Resources
Along with deploying recent manufacturing techniques, the U.S. must update its overall approach. This implies moving away from a model of heavily offshoring production to a small handful of multi-billion dollar foundries, to leveraging the nation’s comprehensive and wealthy arsenal of leading universities, startups and industrial R&D firms to collaborate, speed up discovery and support your entire ‘lab-to-fab’ process (research, prototyping and manufacturing). This will all be achieved while keeping costs in check and integrating enabling technologies directly into these organizations’ infrastructures.
Looking Ahead
The connection between AI and semiconductors is really symbiotic. As we’ve mentioned, semiconductors are crucial for powering the servers that train AI models; on the flipside, AI is significantly accelerating semiconductor materials discovery by leveraging machine learning to predict the properties of latest materials and speed up the design process. This approach, often known as inverse materials design, allows researchers to design materials with specific targeted properties, similar to improved conductivity, energy efficiency and sustainability.
Accelerating the invention of latest materials stays one in every of the hardest challenges in semiconductor manufacturing, though it’s particularly demanding for AI semiconductors, because the industry seeks to continuously drive up computational power, efficiency and speed, while reducing chip size.
While AI will be used to predict the properties of latest, theoretical materials, these breakthroughs have traditionally still been limited by the slow pace of physical validation. Recent techniques will be used to support high-throughput experimentation, helping to shut the gap; enabling faster, more targeted materials development, and ultimately unlocking the subsequent generation of materials. Combining recent techniques like direct atomic layer processing with the facility of AI has the facility to make magic, dramatically accelerating the event of breakthroughs that were never before considered possible, all centralized throughout the U.S.’ own national borders.
