Strategies to Enable Assured Access to Semiconductors for the Department of Defense (2024)
Chapter: 5 Challenges for the Department of Defense in Supporting Sustainability in the Semiconductor Ecosystem
5
Challenges for the Department of Defense in Supporting Sustainability in the Semiconductor Ecosystem
BACKGROUND AND CONTEXT
The Department of Defense (DoD) has distinct electronics needs, particularly when it comes to electronics closely related to the physical layer, the portion of the system closest to the physical world such as close to an antenna or sensor. In domains such as optics, DoD uses different bands and frequencies than typical commercial applications. For applications such as radar, communications, and electronic warfare systems, DoD operates with unique frequencies and power levels. Meeting these niche requirements has required the development of new materials and manufacturing processes for semiconductors tailored to these specific needs.
DoD has been successful, to a significant extent, in fostering a community of firms that can cater to a series of these unique demands. In the realm of compound semiconductors, for example, numerous options are available, typically based on a 6-inch wafer production line. DoD has been able to support and cultivate this manufacturing at a cost and scale that aligns with the needs of these firms, ensuring they possess distinctive capabilities. Over the past few decades, this has been at the core of enhancing warfighting capabilities.
The United States has repeatedly pioneered new materials and integrated them into semiconductor production, leading to technologically advanced warfighting capabilities. These advancements have allowed DoD to detect adversaries before being
detected, extend the reach and range of satellites, and gain a superior understanding of the battlefield compared with their adversaries. Work on integration of “exotic” materials typically takes a decade from invention to end-state manufacturing, but the investment has proven worthwhile in enabling new capabilities. In these activities, DoD has demonstrated that it can establish and sustain a thriving ecosystem. However, as systems have become increasingly digital in nature, the portion of such systems where DoD maintains this advantage has become progressively smaller.
As discussed in Chapter 3, DoD currently faces two challenges. The first is that it primarily operates in a low-volume, high-mix electronics environment, which is quite different from the high-scale production seen in consumer electronics. Consumer electronics thrive on producing millions of components, sometimes costing up to a billion dollars per chip development cycle. To make this investment worthwhile, the market opportunity for these chips usually needs to be very large. As technology has advanced, this scale requirement has become increasingly at odds with DoD’s role in the electronics community. Although this issue is not unique to DoD, it has been particularly pronounced over the past two decades. Efforts to aggregate demand and change this dynamic have largely been unsuccessful. Even increasing purchases from 10,000 to 100,000 units falls short of the scale needed for modern development cycles, and unique requirements often remain unmet. DoD needs to accept and work within this low-volume, high-mix context. DoD has struggled to keep up with the complexities and costs of designing and manufacturing advanced technologies, where scalability is key (Box 5-1). These challenges stem from the nature of low-volume, high-mix markets and are not owing to poor decision-making or lack of expertise. They are common in both commercial and military sectors, which have different scalability factors.
A second challenge is DoD’s diminished leadership in electronics research and development (R&D). In the past, DoD was closely involved in foundational R&D work such as finding the next generation transistor, such as the fin field-effect transistor (FinFET), but more recently it has intentionally focused on security or other more niche, DoD-specific programs, rather than driving the research agenda for the semiconductor sector. Nowadays, commercial factors drive modern semiconductor R&D, which is likely why the Department of Commerce (DOC) has been tasked with creating robust manufacturing operations in the United States through the CHIPS and Science Act of 2022 (CHIPS Act). DoD now needs to adjust to utilizing commercially focused manufacturing facilities without imposing barriers that could hinder U.S. manufacturing competitiveness on a global scale.
There is some alignment between the needs of DoD and the commercial sector in processing and storing large volumes of data, offering potential collaboration opportunities. However, certain policy decisions have complicated DoD’s situation. For example, the Trusted Foundry program, designed to produce microelectronic components while maintaining security, has limited DoD’s partnership opportunities. The Trusted Foundry program’s manufacturing locations do not align with
BOX 5-1
The Unique Department of Defense Challenge for Artificial Intelligence
In the global competition surrounding the use of artificial intelligence (AI), the entity that controls the AI technology stack and supply chain will have a significant advantage in future conflicts. China has been rapidly advancing in developing a native AI technology stack. Recent demonstrations have revealed China’s ability to create cutting-edge semiconductor electronics, surpassing the 10 nm node and producing graphic processing units (GPUs), the specialized chips for AI applications, that rival Nvidia’s Tensor Core A100 class of GPUs for AI, currently a leading AI chip.a
Currently, a significant fraction of the value chain for advanced AI applications originates from U.S.-based companies. This includes the models, cloud infrastructure, accelerators, and electronic design automation (EDA) tools, and most chip manufacturing tools. This advantageous position is acknowledged worldwide, but the Department of Defense (DoD) has not capitalized on this inherent advantage to realize the potential of resulting defense applications.
The Defense Advanced Research Projects Agency (DARPA) recognized the impending AI revolution in 2017 through the AINext program, but instead of spearheading the revolution, DoD chose to study the technology and invest in various small initiatives, which did not lead to an AI infrastructure for defense needs. In contrast, the commercial industry transitioned from studying AI capabilities to productizing them in 2018. For instance, Microsoft’s announcement of an AI supercomputer in 2020 can be traced to decisions made in 2018. This served as the foundation for OpenAI’s ChatGPT capability, which launched in 2023.
Notably, DoD did not develop similar capabilities. To have developed an AI supercomputer, DoD would have needed to start around 2018 when commercial entities were pouring billions of dollars into infrastructure. However, owing to its own regulatory constraints around security requirements, creating its own chips would have required using the 90 nm node, which was the most advanced node supported by the Trusted Foundry program at the time. This is a concern, especially in light of recent announcements about Chinese GPUs created by Huawei.b China’s dual alignment of industry with government gives it both commercial scale and military impact.
Fortunately, the U.S. industry maintains a substantial lead in AI, particularly in areas not directly tied to the unique physical layer. Here, the alignment of DoD needs and commercial interests largely overlaps. DoD requires access to AI technologies that are available from the commercial sector while ensuring a reliable supply during times of need.
__________________
a H. Mujtaba, 2023, “Chinese AI Company Claims Huawei AI GPUs Are on Par with NVIDIA A100, Will Compete with GPT-4 LLM in 2024,” August 27, https://wccftech.com/chinese-ai-company-claims-huawei-ai-gpus-on-par-nvidia-a100-compete-gpt-4-2024/amp.
b AI Beat, 2024, “Huawei’s Proprietary AI Chip Could Exceed Nvidia’s A100 GPU,” March 4.
those used by the advanced semiconductor industry, restricting DoD’s access to cutting-edge chips. Furthermore, the sites within the Trusted Foundry program are now far behind the leading edge in providing modern fabrication options, further isolating DoD from industry trends, modern intellectual property, and advanced semiconductor manufacturing.
Beyond manufacturing challenges, designing modern systems is highly complex and resource intensive. Additional bureaucratic layers and unique steps in the
design and manufacturing processes further complicate these efforts. Even with access to free manufacturing in a modern facility, DoD would struggle to create competitive designs using the most modern nodes. Similar issues affect the commercial industry, as designing at scaled nodes is challenging even for the largest entities, creating widespread competitive pressure in the industry.
ADDRESSING RESEARCH AND DEVELOPMENT BARRIERS
Chapter 3 discusses the challenges faced by the semiconductor industry as it approaches the physical limitations of current complementary metal-oxide-semiconductor (CMOS) technology, particularly as node sizes shrink to 3 nm and 2 nm. This situation indicates that the coming shift to post-CMOS technologies might be necessary within the next decade. To address these challenges, the community will need new architectures, materials, packaging, software, algorithms, and other innovative technologies.
Three primary lines of effort are particularly important: developing more efficient architectures and packaging, creating new models of computation, and discovering new materials and devices. The IEEE’s International Roadmap for Devices and Systems (IRDS) has identified several beyond-CMOS approaches, including computational state variability, nonthermal equilibrium systems, novel energy transfer interactions, nanoscale thermal management, sublithographic management, and alternative architectures. Some promising advances are noted in the equipment sector with atomic layer deposition and etching, and in architectural developments such as in-memory processing, Nano/Micro Electro-Mechanical Systems (N/MEMS), magneto-electric spin-orbit (MESO) logic devices, cryogenic electronics, structures based on magnetic tunnel junctions, and edge computing. Additionally, three-dimensional (3D) heterogeneous integration packaging, which stacks multiple integrated circuits and connects them vertically for improved performance, lower power consumption, and reduced size, is currently receiving significant attention.
DoD is particularly interested in these advancements owing to their importance for national security as well as for economic security. DoD could play a pivotal role in coordinating and supporting next-generation R&D efforts in this domain. This is especially relevant as the funding for the proposed National Semiconductor Technology Center (NSTC) will conclude after 5 years. During this period, DoD should collaborate closely with the National Institute of Standards and Technology (NIST) on NSTC efforts and consider strategies for post-CHIPS Act funding.
An ongoing focus on post-CMOS R&D seems essential and DoD should fully partner on this work with NSTC in the near-term, while also developing its own long-term focus through DARPA’s Electronics Resurgence Initiative (ERI) 2.0 and similar efforts, and while partnering with other agencies to ensure that these R&D efforts continue after the initial 5 years of CHIPS Act funding expires.
Near-Term Advances: Packaging, Integration, and Other Technologies
Recent months have seen several new initiatives in the semiconductor sector, particularly focusing on packaging and chip integration technologies, which are expected to materialize within 5 years. At the forefront of these efforts are DARPA and NIST. DARPA recently announced the establishment of a Next-Generation Microelectronics Manufacturing (NGMM) center with a budget of $430 million. This center is focused on 3D heterogeneous integration (3DHI). This approach marks a shift from traditional monolithic chip production to a more modular method using chiplets that can be assembled like LEGO blocks. This method offers the potential to disaggregate functions such as memory and processing, thereby significantly boosting performance and lowering cost. DARPA’s initiative also includes exploring the integration of photonics and non-silicon electronics into these systems.
Simultaneously, NIST announced the National Advanced Packaging Manufacturing Program (NAPMP), backed by a substantial $3 billion from the CHIPS Act. This program is another major initiative in packaging and integration, covering various elements like materials, tools, thermal management, photonics, and the chiplet ecosystem. The overlapping elements of DARPA and NIST’s programs highlight the need for coordination between the two to maximize efficiency and avoid duplicative efforts.
NIST has established the NSTC as a part of the CHIPS Research and Development Office. The NSTC, conceived as a public–private consortium for semiconductor R&D, will involve academia and industry in addressing industry barriers, including workforce development. While funded for a multiyear period, there is uncertainty about the extension of this funding beyond its initial timeline.
DoD has launched the Microelectronics Commons (ME Commons), a 5-year PPP with industry and university hubs, focusing on six key semiconductor technology areas, including broadband technology, artificial intelligence (AI), and quantum technology—funded by the CHIPS Act. The goal of the ME Commons is to bridge the gap between university research and semiconductor production and to enhance U.S. leadership in these fields. Notably, however, this initiative does not address the need for groundbreaking technologies beyond the CMOS era.
Effective coordination among these initiatives is necessary to ensure the seamless exchange of ideas, prevent duplication, and avoid research dead ends, thereby ensuring optimal advancement in the semiconductor field.
Finding 5.1: Four major public–private partnership (PPP) programs have been formed to date by DoD and NIST in semiconductor R&D and in packaging: the Next-Generation Microelectronics Manufacturing program by DARPA, the ME Commons by DoD, NSTC by NIST, and NAPMP by NIST. Given the importance of research advances to retention of a technology leadership role
by the United States, it is imperative that the agencies and programs cooperate closely on these programs. Other countries have shown that PPPs work best when they are truly national scale efforts, such as IMEC in Europe or ITRI in Taiwan. The convening authority and ability to work with industry is much simpler when there are fewer major efforts, as opposed to multiple smaller efforts. This has been shown by the international success. Moving forward, there should be fewer efforts and national centers, with perhaps a drive toward only one.
Recommendation 5.1: The Department of Defense (through the director of the CHIPS Coordination Cell) should coordinate with the National Institute of Standards and Technology on its semiconductor-related public–private partnership efforts, including via the Microelectronics Commons, to avoid duplication and ensure the exchange of promising advances.
This coordination might involve merging activities and funding vehicles to avoid duplication and ensure exchange of promising advances. It could also involve personnel exchanges and having laboratory facilities that are open for co-work to reduce barriers to access.
For significant progress to be made in the coming half decade on the initial range of semiconductor technologies, starting with advances on packaging and integration, it will be important to assure that researchers are sharing progress and findings. The four DoD and NIST programs should also assist in lowering another barrier to progress in semiconductor research—the need to provide university semiconductor researchers as well as small and medium-sized companies with access to advanced fabrication facilities.
Finding 5.2: University researchers and start-up companies can play significant roles in supporting semiconductor research advances. However, because semiconductor production facilities, particularly advanced facilities, are so expensive, these institutions lack access to the relevant equipment to undertake research as well as related development, prototyping, and testing. Ongoing and newly created DoD and NIST programs and initiatives offer an opportunity to significantly improve the equipment and technology access problems faced by university researchers and innovative new companies.
There are several ways DoD and NIST can implement collaborations. For example, DoD program managers could be incentivized to use the NSTC facilities that are being stood up as part of the CHIPS Act. They could run programs specifically to use the new facilities that are brought online, potentially even with a reduced rate of operations. To ensure university participation, there could
be one or more national conferences that jointly shares results. The CHIPS Act coordination team could review the findings to look for overlaps and promote collaboration. Lastly, there could be budget sharing and interchange of personnel through NATCAST.
Recommendation 5.2: The Department of Defense (through the Office of the Under Secretary of Defense for Research and Engineering and the Office of the Under Secretary of Defense for Acquisition and Sustainment) should actively leverage its own programs and partner with the National Institute of Standards and Technology to connect academic researchers with industry partners and advanced fabrication and packaging facilities.
Post-Complementary Metal-Oxide-Semiconductor Advances
The discussion above highlights DoD’s need to focus on long-term advancements in semiconductor technology, particularly beyond the current CMOS technologies. This shift is crucial to achieving significant gains in speed, energy use, and density in semiconductor devices. The International Roadmap for Devices and Systems (IRDS) of the semiconductor sector suggests that the industry’s future lies in moving beyond CMOS technologies, and there is a pressing need for systematic efforts to ensure these new technologies are developed and accessible within the next decade, as current scaling methods are expected to reach their limits.
A wide range of CMOS approaches offer possible advancements. Researchers have the flexibility to explore various options to find promising solutions. The Nanoelectronics Research Initiative (NRI) of the Semiconductor Research Corporation (SRC) is engaged in benchmarking these potential approaches against one another to assess viability. Besides hardware innovations, significant efficiency improvements can also be achieved through software performance engineering. This involves restructuring software to enhance computer application speed by removing unnecessary elements and optimizing software for specific hardware features. Additionally, advancements in algorithms that require lower computation levels and hardware streamlining that uses fewer transistors and chip area are other avenues for gains.
While NSTC and the ME Commons may include research in some of these areas, there is need for a more systematic and accelerated research effort in post-CMOS technologies. DoD’s role is not just to identify and fund research in these technologies, but also to actively participate in a broader, collaborative research agenda with the industry and other entities like NSTC. The scale of this challenge is beyond what DoD can address alone, underscoring the importance of collaborative efforts in advancing post-CMOS research.
Finding 5.3: As semiconductor fabrication continues to scale ever smaller to achieve speed and energy advances, the physics of CMOS technology is becoming increasingly challenging. In approximately a decade, in order to sustain semiconductor advances to meet the demands of AI, quantum, and more-specialized chips, there will be a need for post-CMOS technologies that advance speed and lower energy consumption through new materials, new architectures, and new computational approaches. DoD will need to take advantage of these post-CMOS R&D initiatives to sustain competitive advantage.
In its quest to assure access to the most advanced manufacturing technology, DoD, via the Office of the Under Secretary of Defense for Research and Engineering (USD R&E), will need to engage in longer-term, early-TRL R&D as well as support device, chip, and package prototyping. This work will be most successful if DoD is a robust partner to NSTC, with sustained investments and clear strategic directions. DARPA’s ERI 2.0 could be leveraged strongly to engage industry and university researchers in this work. Similarly, entities like DoD’s Office of Strategic Capital, the Defense Innovation Unit, and In-Q-Tel could be leveraged strongly for prototyping and scale-up support.
Recommendation 5.3: The Department of Defense (through the Office of the Under Secretary of Defense for Research and Engineering) should focus on long-term scientific research in disruptive semiconductor technologies, including post-CMOS technologies, and ensure broader access to prototyping facilities for academic researchers and small to medium-sized firms.
ARTIFICIAL INTELLIGENCE ADVANCES
It is difficult to overstate the potential value of emerging AI and machine learning (ML) technologies for military use. As noted throughout the report, AI semiconductors are critical for maintaining technological superiority in defense capabilities, enabling advanced systems for national security, cyber defense, and autonomous military operations. Ensuring the security and innovation of these semiconductors is essential for safeguarding sensitive information and enhancing the strategic advantage of the U.S. military. It is very likely that the nations that dominate the AI landscape will also have global national security leadership.
Finding 5.4: In the global competition surrounding the use of AI, the entity that controls its AI hardware and software supply chain will have a significant advantage in future conflicts. While DoD anticipated the arrival of the AI revolution, the necessary investments to create the hardware and software required
were made by the private sector. Today, DoD requires access to AI technologies while ensuring reliable and resilient supply chains.
Recommendation5.4: The Department of Defense (DoD) (through the Chief Digital and Artificial Intelligence Office) should swiftly incorporate U.S.-origin commercially developed artificial intelligence (AI) models, chips, and infrastructure, and develop custom AI technologies for defense purposes, to maintain leadership in this transformative technology. DoD should also track leading-edge AI technologies and align DoD efforts with commercial interests to mitigate supply chain risks.
ADDRESSING DESIGN BARRIERS
The challenge of high costs of designing application-specific integrated circuits (ASICs), which are crucial for achieving technological superiority in military systems, has only grown more acute over time, and it requires the development of new design tools to enable the creation of affordable, high-performance, low-volume custom ASICs (see Chapter 2). Although the commercial electronic design automation industry is developing such tools, the rate of investment is insufficient to meet DoD’s specific needs. Advancements in software and AI technology offer the potential for rapid progress in this area.
The CHIPS Act, focused on production facilities, did not address the critical issue of design costs, and reducing these costs remains a vital barrier to exploiting technological opportunities afforded by access to advanced chip manufacturing technology. This challenge is not unique to DoD, but it is also faced by start-ups, small and midsized firms, and large companies seeking to develop applications for moderate-sized markets using advanced chips. Therefore, a collaborative R&D initiative led by DoD, working with industry to support new lower-cost design approaches, would have significant benefits for both the defense sector and the wider industry.
Human resource constraints are another factor increasing costs and limiting the integration of advanced semiconductors into new DoD systems. The application of AI and ML software tools may reduce the need for large teams of software engineers to do chip design and development, and it may also reduce the (currently steep) barriers to entry into this discipline, thus addressing these constraints.1 This approach is further supported by recent DARPA semiconductor research and prototyping initiatives that include a focus on design automation and simulation software, priming the conditions for a larger opportunity that translates these concepts into practice.
___________________
1 E. Sperling, 2024, “Preparing for an AI-Driven Future in Chips,” Semiconductor Engineering, February 1, https://semiengineering.com/preparing-for-an-ai-driven-future-in-chips.
Finding 5.5: With DoD requiring custom chips for its specialized needs, with the cost of chip design accelerating, and with the need for specialized chips to meet different tasks (such as for AI and quantum) growing, DoD is facing design costs that could affect its security missions. Major segments of industry also have an interest in lowering design costs. Development of AI and other tools may in turn enable a new suite of lower-cost design options and approaches. Therefore, a flagship program to significantly reduce design costs appears in order. If the United States does not undertake this, it could cede its design leadership to others ready to undertake it. An investment in development and deployment of more advanced chip design software tools is a natural and logical complement to DoD’s current ME Commons initiative.
Recommendation5.5: The Department of Defense (through the Office of the Under Secretary of Defense for Research and Engineering) should organize a significant initiative, in partnership with industry, to leverage artificial intelligence and machine learning tools to substantially reduce the time and cost of application-specific integrated circuit design and software development for defense needs.
Recommendation 5.6: The Department of Defense (through the Office of the Under Secretary of Defense for Research and Engineering) should build skills internally for modern chip design, partner with commercial-sector experts designing chips at advanced nodes, and tap into the Department of Commerce–supported supply chains for manufacturing necessary custom chips.
Recommendation 5.7: The Department of Defense (DoD) (through the Office of the Under Secretary of Defense for Research and Engineering) should create a flagship unit with expertise in hardware cybersecurity and modern chip design that DoD program managers can access as needed. DoD should explore embedding design teams from this unit within companies developing leading-edge chip designs to increase DoD’s technical expertise.
ADDRESSING BUREAUCRATIC AND REGULATORY BARRIERS
DoD is facing significant challenges in adopting advanced chip technologies, primarily owing to internal security requirements and the reliance on trusted foundries, as well as regulations like ITAR (International Traffic in Arms Regulations) and EAR (Export Administration Regulations). Complying with these requirements requires a sophisticated understanding of complicated regulations that are administered by multiple agencies and that change frequently. Export controls
are administered by DOC and its Bureau of Industry and Security (BIS), which regulate the export, reexport, and in-country transfer of items on the Commerce Control List (most commercial and some military goods). This includes dual-use items that have civil applications as well as military applications. Before exporting an item, a company must determine whether a license is required based on the nature of the item, its destination (both geographic location and identity of the recipient), and its intended use. ITAR, on the other hand, is administered by the Department of State’s Directorate of Defense Trade Controls (DDTC) and regulates the export of defense items, including data relating to defense items, and defense services. The U.S. Munitions List (22 CFR 121.1) sets out those items subject to ITAR, but it includes numerous ambiguities and catch-alls, often making it difficult to interpret. A registered company must obtain a license or written agreement from DDTC before exporting any ITAR-controlled items or services. Businesses are expected to determine how items should be classified, but failure to comply with ITAR and EAR carries severe penalties, including fines, imprisonment, and debarment. If an exporter has doubt about the proper classification of an item, it must submit a written request called a Commodity Jurisdiction, and DDTC will provide advice on whether something is a dual-use item subject to EAR or a defense item regulated by ITAR. Further complicating the situation, in October 2022 the Commerce Department amended EAR by announcing new regulations targeting China’s ability to acquire semiconductor manufacturing technologies and advanced computer chips. Those controls were expanded in December 2023, further limiting exports of advanced computer chips and associated manufacturing technologies to China and to Chinese entities located in other countries.
This system has led to inefficiencies in acquiring custom advanced chips, impacting national security readiness. The recent development of a new jam-resistant GPS capability by the Air Force, a billion-dollar project, illustrates the regulatory hurdles in using advanced chips. The project design was completed in a short time, but navigating through regulatory barriers to justify the use of U.S. facilities took years.
The Microelectronics Quantifiable Assurance (MQA) system discussed in Chapter 3 aims to assure the confidentiality, integrity, and availability of microelectronics while accessing the commercial supply chain. However, transitioning from the trusted foundry model to an evidence-based assurance approach has been behind schedule owing to various challenges, including difficulties in developing and staffing new processes and the impact of coronavirus disease 2019 (SARS-CoV-2). This delay has put DoD 3 years behind the requirements set by the fiscal year (FY) 2020 NDAA to establish trusted supply chain and operational security standards. DoD faces a critical trade-off: to secure access to the latest microelectronics, it must alter its approach to managing a “secure” supply chain and align more closely with commercial industry practices. DoD should do whatever it can
to align with domestic manufacturing of components, and leverage DOC’s investment to increase resilience in supply chains.
Finding 5.6: DoD requires access to the most advanced chips to keep up with technology advances it must have to meet national security needs. To access cutting-edge chips for its customized needs, it will need to rely on commercial firms, and develop an evidence-based chip quality and security assurance system for a trusted supplier ecosystem. There are significant barriers to accessing the newest U.S.-based plants that will be funded by DOC, based on current DoD rules. This mismatch between security concerns, DoD’s semiconductor needs, and economic resilience is a weakness within the country and stops the United States from having world-class national security efforts.
Recommendation 5.8: The Department of Defense (DoD) (through the Office of the Under Secretary of Defense for Acquisition and Sustainment) should accelerate efforts to implement an evidence-based assurance system to ensure fast and secure access by DoD programs to advanced commercial semiconductors, and also simplify bureaucratic processes and update relevant DoD instructions and policies in support of this effort.
This should be a major, high-priority effort, with appropriate resources allocated to develop and enhance a Microelectronics Evidence-Based Assurance System for secure semiconductors. This system should guarantee DoD’s ability to rapidly acquire advanced, commercially sourced, customized chips for its systems. Once established, program managers for technology projects should have the authority to obtain chips without further review, expediting the integration of advanced chips into systems. DoD should fully implement the requirements of Section 224 of the National Defense Authorization Act for Fiscal Year 2020, while ensuring alignment with commercial practices to avoid creating unnecessary gaps with manufacturers. This includes (1) collaborating with commercial process design kits wherever possible, instead of establishing a separate set of standards; (2) updating DoD instruction 5200.44; (3) publishing a new policy to implement the evidence-based assurance method; (4) updating the Joint Federated Assurance Center Charter and Concept of Operations to develop a process for prioritizing the evidence-based assurance efforts of supporting DoD laboratories; and (5) identifying the resources required to support the National Security Agency’s role in threat analysis for evidence-based assurance, or identifying another DoD organization capable of performing this role. While these have been studied or implemented before, the current focus should be on simplifying, minimizing bureaucratic hurdles, and maximizing access to the supply chain created by the CHIPS Act funding. This alignment of efforts between DOC and DoD is crucial.
Recommendation 5.9: The Department of Defense (DoD) (through the Office of the Under Secretary of Defense for Acquisition and Sustainment) should review and update policies that limit it from manufacturing in commercial facilities. DoD should also increase the use of domestically manufactured chips where possible and simplify procurement processes to streamline access to semiconductor suppliers.
Finding 5.7: Current ITAR and EAR regulate the export of defense-related technologies, information, and services, including information conveyed to (and therefore work performed by) nonpermanent residents. Given the international nature of the semiconductor sector professional workforce, these regulations, as currently interpreted, are limiting DoD’s access to an important talent pool. Accordingly, a new strategy is needed. In addition, there are signals that application of environmental reviews, such as through the National Environmental Policy Act may cause significant construction delays in new fabrication plants (fabs). Further legislative authority may be required to allow DoD to better “onshore” and “friend-shore” required manufacturing activity.
Recommendation 5.10: The Department of Defense (DoD) (through the Defense Technology Security Administration) should collaborate with relevant organizations to reduce the administrative burdens and improve the timeliness of decisions related to International Traffic in Arms Regulations, Export Administration Regulations, and the National Environmental Policy Act. DoD should consider forming a task force to review and reform these regulations.
ADDRESSING BARRIERS TO MANAGING INTELLECTUAL PROPERTY RIGHTS
Effective management of intellectual property (IP) rights presents another challenge that DoD must resolve in a PPP, including not just patent rights, but also data rights in computer software and technical data, mask works, and trade secrets and related know-how. Consideration must be given to ownership of IP created during the PPP, government rights to IP created with the direct or indirect use of government funding, requirements for sharing or transferring IP to other PPP participants and/or to the public, and restrictions on publication and other dissemination of information. Four general models can be utilized by DoD, depending on the technology readiness level (TRL) and intended use: (1) open access to IP; (2) sharing IP between members of the PPP only; (3) retention of IP rights by each PPP to IP created by the participant; and (4) when necessary, federal government ownership of IP created during the PPP. Each is discussed in turn below.
Open Access to Intellectual Property
PPPs are sometimes characterized as a basic form of institutionalized knowledge sharing, directed to creating a “commons” where information is collectively owned and managed. Other commentators have argued that PPPs should recognize a form of patent fair use in order to promote collaborative research.2 However, in the context of a PPP intended to strengthen DoD supply chain for semiconductor devices, open access to IP has the potential to dramatically undermine the very purpose of the PPP and could potentially lead to sharing of information with adversaries. Instead, DoD must create a model that preserves the rights of private partners and their ability to commercialize inventions created in the PPP.
Open access to the general knowledge and basic skills of industry workers (the level of ordinary skill in the art) requires a different analysis, because such knowledge and skills are likely already in the public domain. One area where open IP has been touted is in the context of workforce training programs. Intel has created Semiconductor Education and Research Programs with a group of Ohio universities and community colleges, which it describes as an “open IP” model.3 In some instances, Intel Laboratories conducts its own research first, then explores ways in which that research can be augmented by graduate students working at several universities (known as the Academic Mindshare program). Another effort will provide experiential opportunities for students, including a 1-year certificate program for technicians that makes them eligible for entry-level positions at Intel or another semiconductor company. In these settings where little or no IP will be created and the focus is on training technicians and other lower-level employees, the concept of open IP takes on a different meaning.4
Consortium Models for Sharing Information
The Semiconductor Manufacturing Technology consortium (SEMATECH) is one example of a consortium where IP was shared between its members and DoD (see Box 5-2). It was considered to be an extremely successful consortium for developing advanced manufacturing technology, due at least in part to the sharing by member companies of precompetitive information about their manufacturing
___________________
2 See, for example, L.S. Vertinskey, 2015, “Patents, Partnerships, and the Pre-Competitive Collaboration Myth in Pharmaceutical Innovation,” UC Davis Law Review 48:1509 (writing in the context of pharmaceutical PPPs), and other references cited therein.
3 S. Venkataramani, Intel Labs, presentation on November 28, 2023.
4 Ms. Venkataramani noted that sponsored research agreements are available if needed.
BOX 5-2
SEMATECH
SEMATECH, Inc., was formed in August 1987 as a nonprofit corporation to conduct research and development intended to provide the U.S. semiconductor industry with domestic capabilities to position it as a world leader in semiconductor manufacturing. It was a consortium of 14 major U.S. semiconductor and computer companies, where each company assigned technical employees (“assignees”) to participate in SEMATECH’s research and development (R&D) programs.a The memorandum of understanding (MOU) between SEMATECH and the Department of Defense (DoD) contained two separate terms addressing ownership and use of intellectual property.
First, the MOU required SEMATECH to take all steps necessary to maximize the expeditious and timely transfer of technology developed and owned by SEMATECH to its participants, in accordance with the agreement between SEMATECH and its participants.b As a result, inventions created by the assignees were patented by SEMATECH, and then transferred to participants.c Second, even though DoD was not a formal member, apparently because of its financial contributions, the MOU provided that DoD was permitted to use intellectual property, trade secrets, and technical data owned and developed by SEMATECH in the same manner as a participant in the consortium. DoD was also permitted to transfer such intellectual property, trade secrets, and technical data to DoD contractors for use in connection with DoD requirements, provided that DoD did not transfer IP to any person for commercial use.d
__________________
a Each company signed a participation agreement requiring the company to support SEMATECH at certain funding levels for an initial period of 4 years. A member company’s annual financial contributions were calculated based on its prior year’s sales of semiconductor devices (for those members who produced electronic equipment, the value of its semiconductor purchases in the previous year was used instead). A company’s financial contributions then determined the number of technical employees it could assign to SEMATECH, along with the extent of technology and associated know-how that could be transferred to the company. See U.S. General Accounting Office, 1989, “The SEMATECH Consortium’s Start-up Activities,” November, p. 11.
b SEMATECH, 1988, “Memorandum of Understanding,” May 12, p. 4.
c P.N. Dunn, 1997, “SEMATECH’s Burgeoning Patents May Be Future Revenue Generator,” Solid State Technology 40(1).
d SEMATECH, 1988, “Memorandum of Understanding,” May 12, pp. 3-4. See also 15 U.S.C. § 4602(b).
processes and equipment.5 It is unclear, however, whether this model would attract participation today for anything beyond basic research and pre-commercialization efforts (TRL 1–3). Companies appear less interested in working collaboratively because they want to preserve their IP rights to both background IP and new developments, and they frequently prioritize profits over the collective good of the
___________________
5 U.S. General Accounting Office, 1992, “Lessons Learned from SEMATECH,” September, p. 4. The GAO report further discussed the importance of SEMATECH as a forum for communications between members. Specifically, through SEMATECH discussions, industry members found that they were protecting similar trade secrets and attempting to address similar manufacturing problems. Id. at p. 8.
industry or the needs of DoD.6 Start-ups and small firms, on the other hand, may show more willingness to collaborate and share developments, particularly if they can license background IP that is necessary to their R&D efforts.
Retention of Intellectual Property Rights by Public–Private Partnership Participants
In order for any PPP to be successful, it must allow for private ownership of any IP developed during the partnership that relates to advanced TRLs. In a typical situation where one private party contracts with a second private party for research services, the paying party expects to own any IP created as a result of the relationship. In dealing with the government, however, the relationship is inverted: while the government provides funding, the private party conducting the research expects to own any resulting IP. Daniel Armbrust, former president and chief executive officer of SEMATECH, acknowledged that ownership of IP is critical, particularly for start-up companies. Indeed, he described it as their “lifeblood.”7 Private ownership of IP is recognized and endorsed by the Bayh-Dole Act and federal regulations governing data rights.
The federal Bayh-Dole Act8 applies to inventions arising from federally supported R&D and specifically authorizes private parties to own patents resulting from such work. The act expresses the policy and objective of Congress to, inter alia, promote the commercialization of and public availability of inventions made in the United States by U.S. industry and labor, as well as to ensure that the government obtains sufficient rights in federally supported inventions to meet the needs of the government and protect the public against nonuse or unreasonable use of such inventions.9 The act originally applied only to small businesses, universities, and other nonprofit organizations; however, in 1983, President Reagan
___________________
6 That concern is reinforced by the experience of IMEC, a nanoelectronics center located in Leuven, Belgium, that has been involved in collaborative R&D for more than 39 years. IMEC serves as a bridge between universities, government, and industry, although it receives 75 percent of its funding from industry sources. Much of IMEC’s work is at the precompetitive stage, where both costs and IP are shared and research results are pooled. An IMEC member company pays a background fee to obtain IP created prior to joining, followed by annual fees for access to future developments. See L. Lauwers, IMEC, presentation to the committee, July 18, 2023. Mr. Lauwers explained that IMEC also provides research services at more competitive stages, where IP can be owned exclusively by the company paying the costs and fees associated with the work. This is separate from the consortium program.
7 See D. Armbrust, Silicon Catalyst, presentation to the committee, September 26, 2023. Mr. Armbrust suggested that having an IP template for a PPP is unlikely to be successful, however. He observed that every situation is different: each one has novel issues specific to the stakeholder, funding, and technologies involved. Standard IP terms are therefore not likely to be useful, and flexibility is required.
8 35 U.S.C. § 200 et seq., enacted 1980.
9 35 U.S.C. § 200.
extended its benefits to all contractors, including large businesses and for-profit organizations.10
The Bayh-Dole Act provides that a party to a funding agreement may, within a reasonable time after disclosure of a subject invention, elect to retain title to that invention.11 In order to invoke the process, the contractor must disclose a subject invention to the federal agency within a reasonable time after it becomes known to contractor personnel responsible for the administration of patent matters. The contractor must then make a written election to retain title within 2 years after disclosure to the federal agency, except in those cases where the inventor’s 1-year filing window under Section 102(b) would end before the end of the 2-year election period. The contractor must then file a patent application prior to the expiration of the 1-year period allowed by Section 102(b) and then file corresponding patent applications in other countries in which it wishes to retain title. The federal government may receive title to any subject inventions in the United States or in other countries in which the contractor has not filed patent applications on the subject invention within such times.12
However, even when a contractor elects to retain title to a subject invention, the federal government is not devoid of rights in the invention it funded. Instead, the federal agency receives a nonexclusive, nontransferable, irrevocable, paid-up license to practice or have practiced for or on behalf of the United States any subject invention throughout the world.13 For example, in an instance where DoD funds the development of a new ASIC, its license rights allow DoD to share the subject IP with another contractor for purposes of having the second contractor make the same ASIC for DoD. DoD should not be required to pay twice for the creation of the same invention.14 While DoD is sometimes forced to pay twice for certain chips
___________________
10 See Presidential Memorandum dated February 18, 1983. An exception can be made where (1) it is necessary to obtain an agreement with a uniquely qualified contractor or (2) the award involves co-sponsored, cost sharing, or joint venture research and development and the performer, co-sponsor, or joint venturer is making substantial contribution of funds, facilities, or equipment to the work performed under the award.
11 35 U.S.C. § 202(a). The term “subject invention” means any invention of a contractor conceived or first actually reduced to practice in the performance of work under a funding agreement. The term “funding agreement” means any contract, grant, or cooperative agreement entered into between any federal agency and any contractor for the performance of experimental, developmental, or research work funded in whole or in part by the federal government. See 35 U.S.C. § 201 (definitions).
12 35 U.S.C. § 202(c).
13 35 U.S.C. § 202(c)(4). The government also has the ability to order that an invention be kept secret and can withhold publication of an application or the grant of a patent, when the government believes that disclosure could be detrimental to national security. See 35 U.S.C. § 181.
14 Any U.S. patent that issues must include a statement indicating that the invention was made with government support and that the federal government has certain rights in the invention. See 35 U.S.C. § 202(c)(6).
or capabilities, that duplication of costs likely does not arise as a result of the first contractor’s election to retain title to a federally funded invention. More often, it results from pressure by subsequent contractors to be paid the maximum amount possible to fund creation of a DoD system or device,15 or from DoD’s poor recordkeeping about those inventions in which it has rights.16 DoD apparently does a very poor job of tracking inventions resulting from federal funding, in which it now has nonexclusive rights. It is recommended that DoD conduct a comprehensive audit of patent and other IP rights in which it may presently have an interest. In the future, DoD should implement a central recordkeeping system to track invention disclosures, patent applications submitted to the U.S. Patent and Trademark Office and to foreign patent offices, and issued patents resulting from federal funding.
To the extent practicable, federally funded IP should be manufactured in the United States. The Bayh-Dole Act expresses a strong preference for U.S. industry. A business that receives title to a federally funded invention is precluded from giving an exclusive right to use or sell the invention in the United States unless the products will be manufactured substantially in the United States, unless domestic manufacture is not commercially feasible.17 Similar requirements apply when the government retains ownership of an invention that it funds. The government should also place appropriate restrictions on the sale of federally funded IP to foreign businesses and foreign investors.
The federal government also retains “march-in” rights in federally funded inventions, allowing it to grant an exclusive or nonexclusive license to a third party when a contractor fails to take steps to achieve practical application of the invention within a reasonable time or where such action is necessary to alleviate health or safety needs that are not being satisfied by the contractor or its licensees.18 To date, the federal government has never exercised its march-in rights.19
___________________
15 But see FAR § 27.306 and DFARS § 252.227-7038(i), regarding rights of the U.S. government in background inventions. The government may claim that its license to use a subject invention has little value, unless it also acquires a license to the preexisting technology.
16 DoD apparently does a very poor job of tracking inventions resulting from federal funding, in which it now has nonexclusive rights. It is recommended that DoD conduct a comprehensive audit of patents and other IP rights in which it may presently have an interest. In future, DoD should implement a central recordkeeping system to track invention disclosures, patent applications submitted to the USPTO and to foreign patent offices, and issue patents resulting from federal funding.
17 See 35 U.S.C. § 204.
18 35 U.S.C. § 203(a) (December 12, 1980; last amended January 4, 2011).
19 Comment of U.S. Federal Trade Commission to the Department of Commerce, National Institute of Standards and Technology, “Draft Interagency Guidance Framework for Considering the Exercise of March-in Rights,” at 2 (February 6, 2024), available at Comment of the United States Federal Trade Commission (ftc.gov). See also Congressional Research Service, March-In Rights Under the Bayh-Dole Act at 8 (August 22, 2016).
Recently, however, the Biden administration proposed using march-in rights to lower drug costs, and NIST issued a request for comments on use of march-in rights and forced licensing more generally.20
Ownership and use of data rights must also be addressed, because they apply to both trade secret information and works subject to copyright. The federal government obtains rights in technical data, including a copyright license, under an irrevocable license granted (or obtained for) the government by the contractor. The contractor retains all rights in the data not granted to the government. The scope of the license is generally determined by the source of funds used to develop the data.21 To encourage contractors to offer or use commercial products to satisfy military requirements, contractors shall generally not be required to (1) furnish technical information related to commercial items or processes that is not customarily provided to the public or (2) relinquish to, or otherwise provide, the government rights to use, modify, reproduce, release, perform, display, or disclose technical data pertaining to commercial items or processes except for a transfer of rights mutually agreed upon.22
Accordingly, for commercial items, DoD acquires only the technical data customarily provided to the public with a commercial item or process, with three exceptions. Those exceptions include technical data that (1) are form, fit, or function data; (2) are required for repair or maintenance of commercial items or processes, or for the proper installation, operating, or handling of a commercial item; or (3) describe modifications made at government expense to a commercial item or process in order to meet the requirements of a government solicitation.23
For noncommercial items, DoD acquires only the technical data and rights in that data sufficient to satisfy DoD’s needs.24 The standard license rights that a licensor grants to the government are unlimited rights, government purpose rights, or limited rights.25 Generally speaking, DoD acquires unlimited rights in technical data only when an item, component, or process has been developed exclusively with government funds. DoD acquires only lesser, government purpose rights in technical data relating to an item, component, or process developed with mixed
___________________
20 88 Fed. Reg. 85593 (December 8, 2023). The Federal Trade Commission issued a comment supporting “an expansive and flexible approach” to march-in rights, including use of march-in rights to lower drug prices.
21 DFARS § 227.7103-4(a).
22 DFARS § 227.7102-1(b).
23 DFARS § 227.7102-1(a).
24 DFARS § 227.7103-1.
25 DFARS § 227.7103-5.
funding. For technical data relating to an item, component, or process developed exclusively with private funds, DoD receives only limited purpose rights.26
Government purpose rights are rarely used and present an artificial barrier to working with nontraditional entities. DoD should consider removing government purpose rights for semiconductor engagements and replacing them with a U.S. manufacturing preference. Such a preference is already incorporated in the Bayh-Dole Act, which provides that no small business or nonprofit organization that receives title to a subject invention shall grant any person the exclusive right to use or sell the invention in the United States unless that person agrees that any products embodying the invention will be manufactured substantially in the United States.27
By allowing private parties to retain ownership of IP created during a PPP (including patents, copyrights, and trade secrets), both large and small companies will be incentivized to participate in the partnership because they will emerge with intellectual assets that build value in their companies. Private parties must be able to commercialize the technologies they create in the context of a PPP. The costs of participating in these efforts are so high (both the direct cost sharing and associated operating costs) that private partners must have the ability to use IP created out of a PPP.28 That need is further reinforced by NIST’s request for comments regarding the NSTC, where respondents clearly stated they need to know how IP rights are going to be allocated before they can make any decision about whether to participate in a PPP.29 Remember that corporations have a duty to investors and shareholders to maximize profits.
Government Ownership of Intellectual Property
Government ownership of IP should be restricted to those extreme situations where ASICs are being developed solely for use by DoD and the chips have no commercial application. Because the objective of this report is to consider PPPs that benefit both DoD and dual-use needs, government ownership should arise
___________________
26 DFARS § 227.7103-5. In unusual situations, these standard rights may not satisfy the government’s needs or the government may be willing to accept lesser rights in data in return for other considerations. In those cases, a special license may be negotiated.
27 See 35 U.S.C. § 204. Note though that in individual cases, the requirement can be waived by the federal agency that funded the invention if the patent owner shows that reasonable but unsuccessful efforts were made to grant licenses to persons that would be likely to manufacture substantially in the United States or that under the circumstances domestic manufacture is not commercially feasible.
28 Presentation by Mukesh Kare, IBM, September 5, 2023.
29 NIST, 2022, “Incentives, Infrastructure, and Research and Development Needs to Support a Strong Domestic Semiconductor Industry: Summary of Responses to Request for Information,” Special Publication NIST SP 1282, https://doi.org/10.6028/NIST.SP.1282.
only rarely. In those unusual circumstances, DoD would own some or all of the IP resulting from a PPP, and it could then transfer that technology to government contractors on a limited basis, subject to appropriate security clearances and controls. Government ownership is authorized by the Bayh-Dole Act, which recognizes that in some circumstances a contractor may not have a right to retain title in a subject invention, and the federal government will own inventions made under a federal funding agreement. Government ownership is required where the contractor is not located in the United States, does not have a place of business located in the United States, or is subject to the control of a foreign government. In exceptional circumstances, a funding agreement may also restrict or eliminate the right to retain title when the federal agency determines that such restriction or elimination of rights will better promote the policy and objectives of the act. Similarly, if a federal agency is authorized to conduct foreign intelligence or counterintelligence activities, the right to retain title may be restricted or eliminated as necessary to protect the security of those activities.30 The data rights provisions in the Defense Federal Acquisition Regulation Supplement (DFARS) likewise authorize DoD to acquire unlimited rights in technical data relating to an item, component, or process that has been developed exclusively with government funds.31
Finding 5.8: Consortium models that require broad sharing of background IP and IP developed during a PPP are unlikely to be attractive to industry members beyond TRL 1–3. At more advanced TRLs, industry members will expect to own IP created during the PPP in order to preserve the right to commercialize inventions and maximize profits. Private ownership of IP is expressly authorized by the Bayh-Dole Act and the federal regulations governing data rights.
Recommendation 5.11: The Department of Defense (DoD) (through the Office of the Under Secretary of Defense for Research and Engineering) should manage intellectual property (IP) rights based on technology readiness levels and intended end use, recognize private ownership of IP in the majority of cases, reserve government ownership only for extreme situations, conduct an audit to track DoD’s existing IP rights, and create a central, searchable records system.
___________________
30 35 U.S.C. § 202(a). A final restriction relates to the operation of government-owned, contractor-operated facilities of the Department of Energy primarily dedicated to DOE’s naval nuclear propulsion or weapons related programs.
31 Id. In unusual situations, these standard rights may not satisfy the government’s needs or the government may be willing to accept lesser rights in data in return for other considerations. In those cases, a special license may be negotiated.
ADDRESSING MANUFACTURING BARRIERS
Chapter 3 summarizes the significant challenges in manufacturing processes, especially as production costs have risen sharply owing to the demands of scaling in semiconductor manufacturing. This issue is not unique to DoD but is common across industries and is linked to market size constraints. DoD’s need for ongoing access to leading-edge chips necessitates encouraging advanced manufacturing capabilities within the U.S. industry. While the majority of semiconductor production will remain in Taiwan, South Korea, and other countries, having some capacity in the United States is crucial for DoD’s future technology needs. DoD’s Office of Strategic Capital could become a significant player in the semiconductor ecosystem if it increases its financing level.
With the CHIPS Act being a 5-year authorization and no guarantee of extension, DoD, in collaboration with DOC, needs to consider extending financing mechanisms for advanced semiconductor facilities beyond the act’s timeframe. The issues of U.S. chip leadership and the need for government financing to ensure production capability in the United States are expected to persist beyond this period. Therefore, proactive planning for continued manufacturing support is essential for DoD to maintain access to the advanced semiconductor technologies critical for national defense.
Finding 5.9: U.S. firms have lost leadership in producing the most advanced chips, yet DoD’s future needs for advanced technologies dictate that it has access to at least one cutting-edge semiconductor fab in the United States, as well as other domestically located key elements of the semiconductor ecosystem. Since the CHIPS Act is only a 5-year authorization for support for chip production and DoD’s needs will continue long beyond that term, it is important that financing support tools (e.g., subsidies, loan guarantees, and tax credits) be extended to support advanced semiconductor facility construction in the United States on a continuing basis.
Recommendation 5.12: The Department of Defense (DoD) (through the director of the CHIPS Coordination Cell) should act to secure longer-term federal funding support for the semiconductor sector, beyond the term of the CHIPS and Science Act of 2022, to ensure DoD has continual access to the most advanced semiconductor technologies in the world. DoD should also ensure that these technologies can be sourced from diverse foundries located in the United States and friendly nations.
DoD needs continuous access to advanced semiconductor foundries capable of both high- and low-volume manufacturing to meet its specific requirements, including in emergency situations. However, these foundries, to be commercially
competitive, cannot primarily cater to DoD needs alone. This situation necessitates that DoD adapt its security controls in hardware supply chains and accept some level of risk that comes with commercial partnerships. DoD’s approach should focus on foundries that serve both defense and commercial markets (dual use) that are capable of producing both high and low volumes and maintain a competitive environment with multiple producers in the United States and allied countries. This approach acknowledges the realities of global supply chains and the necessity of international collaboration.
DoD has initiated programs like Rapid Assured Microelectronics Prototypes (RAMP) and RAMP-C to develop secure design and prototyping capabilities without relying on closed security architectures. These programs have enabled partnerships with leading firms like Intel, Boeing, and Northrop Grumman. In addition, DoD should consider PPP models to ensure access to advanced semiconductor manufacturing capacity. These models could include government loan guarantees combined with direct contributions, similar to those used in the private semiconductor industry and outlined in the CHIPS Act.
DoD also needs to prepare for future technology evolution, following the current period of support by the CHIPS Act. This preparation will necessarily involve both R&D and capital investments for leading-edge facilities and ensuring access to these facilities. The federal government’s role in financing advanced fabrication facilities is crucial, as seen from the past two decades of global semiconductor manufacturing.
Finding 5.10: Because DoD has a critical future need for advanced chips, if federal government financing is continued past the 5-year authorization of the CHIPS Act, it will be important that DoD leverage collaborative agreements to assure access for its advanced chip needs as part of such financing.
Recommendation 5.13: The Department of Defense (DoD) (through the director of the CHIPS Coordination Cell) should act to ensure that federal capital and financing supports for the semiconductor sector are tied to agreements that guarantee it access to advanced semiconductors. The DoD Office of Strategic Capital should also prioritize financing for critical technology scale-up activities.
Finding 5.11: DoD has a specific challenge with speed of execution for both design and manufacturing. The Air Force study of the quantifiable assurance development effort32 highlights the substantial gap in clarity that can slow a
___________________
32 V. Coleman, 2023, “NDAA 2023 Mandated Independent Review of USD (R&E) Microelectronics Quantifiable Assurance Effort,” Department of the Air Force, August 3, Version 1.0, https://www.af.mil/Portals/1/documents/2023SAF/MQA_Report.pdf.
program by years. The lack of manufacturing options, and clarity in what is allowed in existing manufacturing options, has added to the burden of making complex chip designs. Layers of decision-making around manufacturing (trusted foundry, ITAR, EAR) have slowed the process of chip design and manufacturing. In lieu of the ability to make its own custom ASICs in a timely manner, DoD often utilizes commercial chips (often created overseas).33
Recommendation 5.14: The Department of Defense (DoD) (through the Office of the Under Secretary of Defense for Acquisition and Sustainment) should lower barriers to utilizing domestic manufacturing entities, minimizing complex bureaucracy, regulations, and requirements, to obtain the custom chips that DoD needs. The process for determining where DoD’s manufacturing can be performed should have short response times so as to avoid unduly hindering DoD’s technology programs.
Finding 5.12: U.S. industry is far ahead of the U.S. government in semiconductor manufacturing. It would be self-defeating for DoD to create a siloed infrastructure for DoD use alone.
Recommendation 5.15: The Department of Defense (DoD) (especially the Chief Digital and Artificial Intelligence Office) should prioritize partnering with industry over creating custom solutions, utilizing commercial technology wherever possible. DoD should partner closely with U.S. companies to creatively and nimbly adopt emerging technologies for defense purposes, with immediate urgent attention to artificial intelligence and potential superintelligent systems.
Only by collaborating with the commercial entities that are creating the most advanced, next-generation, large-scale AI systems will DoD develop the needed understanding to use them as needed, and before adversaries do so. DoD will benefit from partnering with industry leaders to ensure that the best-in-class AI tools are resilient, secure, and composed of U.S.-centric software and hardware, most especially as we head toward creating superintelligent systems.
___________________
33 The preceding paragraph was modified following the release of the report to clarify the specific chips, ASICs, to which it refers.
ADDRESSING THE BREADTH OF THE DEPARTMENT OF DEFENSE’S SEMICONDUCTOR CHALLENGES
As discussed in the preceding chapters, DoD faces significant challenges to develop new microelectronics-based systems, maintain existing electronic technologies across its sprawling enterprise, and remain at the forefront of new semiconductor technologies. Newer organizations such as the Defense Innovation Unit (DIU) have attempted to expedite the adoption of technologies like AI and machine learning by leveraging commercial technologies.
Finding 5.13: DoD’s major defense platforms can last for decades, and the microelectronics require regular updating to remain current and effective. In parallel, DoD needs to be at the forefront of new technology developments in areas such as AI, quantum information science, and advanced data management to remain competitive with other militaries. Clear competitive pressures also accelerate the timelines for DoD to adapt its existing platforms to the latest electronics advances.
Finding 5.14: Also, as DoD works to modernize its various systems, special care is needed to avoid inclusion of fairly mundane legacy semiconductor chips for which the manufacturing may have shifted from allied nations to China during recent years, which could lead to security vulnerabilities. DoD lacks a comprehensive semiconductor strategy to accelerate utilization of the very latest technology in its existing and new systems and platforms.
Recommendation 5.16: The Department of Defense (through the Office of the Under Secretary of Defense for Acquisition and Sustainment, in partnership with the service branches) should develop an overarching microelectronics strategy for research, development, procurement, sustainment, and modernization.
