Access Official Technical PDF

Share Summary:LinkedInEmail

The AI chip supply chain depends on far more than chip designers and fabrication plants. Advanced processors also require specialized materials, manufacturing equipment, software, packaging, and distribution—all supplied through a highly globalized network. A 2019 analysis by the Center for Security and Emerging Technology (CSET) identified photolithography equipment as a particularly important constraint in producing leading-edge chips. For industrial automation leaders, this illustrates how a small number of specialized equipment suppliers can influence the availability of computing hardware used in demanding AI applications.

Why AI chips matter to industrial computing

AI chips are specialized integrated circuits designed to train or run computationally intensive algorithms efficiently. The category includes graphics processing units (GPUs), field-programmable gate arrays (FPGAs), and application-specific integrated circuits (ASICs). These devices support two different stages of AI workloads: training, in which a model is built from data, and inference, in which a trained model processes new inputs.

General-purpose processors can perform many types of computing tasks, but specialized AI chips can offer higher speed and energy efficiency for AI workloads. That distinction matters when an application requires substantial computation or fast execution. The source analysis notes that using less specialized chips can make training substantially more expensive and inference slower—an important consideration for systems that depend on timely results.

In industrial settings, this makes AI hardware a supply-chain issue as well as a computing choice. Access to capable chips depends on whether the interconnected industries that design and manufacture them can deliver the required components and production capacity.

The AI chip supply chain spans multiple industries

No country can produce leading-edge AI chips using only domestic resources, according to the CSET analysis. Production relies on specialized capabilities distributed across several sectors:

  • Research and chip design: Develop new chip architectures and the electronic design automation software used to create them.
  • Manufacturing equipment and materials: Supply the tools and inputs needed to turn a design into a physical device.
  • Fabs: Use equipment and materials to manufacture chips, either for their own designs or as contract manufacturers.
  • Assembly, testing, packaging, and distribution: Prepare manufactured chips for use and move them to customers.

Leading-edge semiconductor manufacturing is particularly concentrated. The analysis attributes this to high technical complexity, significant costs, and economies of scale, which make it difficult for new competitors to enter. As a result, a disruption or restriction affecting one specialized supplier can have consequences beyond that company or country.

Photolithography equipment is a critical bottleneck

Photolithography is a key part of semiconductor fabrication: it is used to define patterns on a wafer during chip production. The CSET paper describes photolithography tools as among the most complex and expensive types of semiconductor manufacturing equipment. It identifies Dutch and Japanese companies as the principal suppliers of these tools, while noting the United States’ role in research and development for advanced photolithography.

The analysis distinguishes between two important technologies. Extreme ultraviolet (EUV) lithography was identified as necessary for manufacturing state-of-the-art 5-nanometer chips and smaller future nodes, though those chips were not yet in mass production at the time of publication. Argon fluoride (ArF) immersion lithography was used across a range of nodes from 45 to 7 nanometers. The paper identified ASML as the sole EUV supplier and as a dominant supplier of ArF immersion tools, with Nikon the other supplier of the latter technology.

Other equipment categories also matter. The paper identifies deposition, etching, and process-control tools—areas in which U.S. and Japanese companies were leading—as secondary constraints. These tools perform different manufacturing and inspection functions, but they are part of the same production ecosystem: restricting access to a key tool can affect the usefulness of complementary equipment and the operation of a fab.

Why equipment access shapes manufacturing capability

Having a fabrication plant does not, by itself, guarantee the ability to produce advanced chips. Fabs need access to the specialized equipment required to manufacture at particular process nodes. The 2019 paper noted that Chinese fabs had reported manufacturing capability as advanced as the 14-nanometer node at that time, but relied on imports from the United States, the Netherlands, and Japan for semiconductor manufacturing equipment.

Its assessment argued that without advanced photolithography and related equipment, domestic production capability would eventually be limited to less advanced nodes. It contrasted that possibility with the paper’s observation that leading-edge AI chips were typically fabricated at 16 nanometers and smaller. These node comparisons reflect the paper’s 2019 analysis and should be read in that historical context, rather than as a description of present-day manufacturing capabilities.

The paper also pointed to barriers beyond equipment purchases. Developing a competitive equipment industry requires technical know-how, skilled labor, capital, and the ability to operate effectively in a highly consolidated market. In complex manufacturing, access to designs or funding alone may not substitute for knowledge acquired through engineering and production experience.

Export controls involve trade-offs

Because advanced lithography suppliers are concentrated in a small number of countries, the paper considered coordinated export controls by the United States, the Netherlands, and Japan as a way to limit access to certain tools. It also emphasized potential costs and unintended consequences. These included reduced revenue for equipment companies, incentives to shift operations to countries outside a control regime, and the risk of retaliation or heightened trade tensions.

Controls on one category of equipment could also affect suppliers of complementary tools if customers could no longer use those tools as part of an advanced production line. This illustrates a broader supply-chain challenge: policy decisions affecting one manufacturing input can alter demand across connected parts of the semiconductor equipment sector.

What this means for industrial decision-makers

The practical lesson is that advanced AI hardware depends on a concentrated, international manufacturing base. For plant managers, system integrators, and technology planners, evaluating AI deployment therefore involves more than comparing chip specifications. It also means recognizing that semiconductor production depends on specialized equipment and suppliers that cannot be quickly replaced.

The CSET analysis highlights photolithography as a strategic point of constraint, alongside the importance of deposition, etching, and process control. It also shows why decisions intended to limit access to manufacturing technology can carry commercial and geopolitical trade-offs. Understanding these dependencies provides a clearer picture of the forces that shape AI chip availability and, in turn, the hardware options available to industrial organizations.

Access Official Technical PDF

About The Author