Executive Summary

• The Guaranteeing Access and Innovation for National Artificial Intelligence (GAIN AI) Act seeks to address domestic supply constraints and inequality of access by mandating U.S. domestic consumers have priority access to advanced semiconductors before international sales can be fulfilled.
• This policy, however, risks unintentionally harming U.S. technological innovation and advantaging foreign competition by broadly limiting access of leading U.S. chip producers to profitable global markets.
• To sustain long-term U.S. AI competitiveness, policymakers must abandon broad restrictions in favor of a strategy focused on establishing narrow, evidence-based export controls limited strictly to clear security threats and accelerating domestic production and infrastructure development.

Introduction

The global AI leadership race is now a race for the most advanced chips. High-performance chips, such as GPUs and other AI accelerators, are the pillar of training and running advanced AI models. Today, U.S. companies – including NVIDIA and AMD – lead in the design and production of these chips. This edge in semiconductor technology is a major part of the United States’ strength in AI and a key piece of its broader geopolitical strategy.

To sustain this competitive edge, the federal government has increasingly deployed export controls as a strategic tool to restrict access to advanced AI chips. The bipartisan Guaranteeing Access and Innovation for National Artificial Intelligence (GAIN AI) Act seeks to address domestic supply constraints and inequality of access by mandating U.S. domestic consumers have priority access to advanced semiconductors before international sales can be fulfilled. It risks unintentionally harming U.S. technological innovation, however, and advantaging foreign competition by broadly limiting access of leading U.S. chipmakers to profitable global markets.

The debate around the GAIN AI Act mirrors past challenges in managing the global spread of advanced AI chips, especially as the government decides how to handle a technology used in both civilian and military contexts. Yet to sustain long-term U.S. AI competitiveness, policymakers must abandon broad restrictions in favor of a strategy focused on establishing narrow, evidence-based export controls limited strictly to clear security threats and accelerating domestic production and infrastructure development.

The GAIN AI Act and the Redefinition of AI Chip Exports

To protect the U.S. technological edge and competitive advantage in artificial intelligence (AI), the federal government has imposed export controls that restrict access to advanced AI chips, particularly to China. Over the past years, successive administrations have introduced restrictions on chip exports. Notably, the first Trump Administration imposed export restrictions on advanced chips to China, including NVIDIA’s A100 and H100, which are designed for high-performance AI. Subsequently, the Biden Administration put in place the Framework for Artificial Intelligence Diffusion, a model that proposed a three-tiered system to control the global trade of U.S. AI chips. (This framework was rescinded by the incoming Trump Administration, citing concerns that such restrictions could ultimately hinder U.S. innovation and leadership in the AI sector.)

The recently introduced bipartisan GAIN AI Act represents a continuation of these export control initiatives and frames the availability of AI chips as a matter of national security and economic stability. The Act would require any U.S. chipmaker to meet a domestic quota before exporting chips. Specifically, it requires a presumption of denial of export licenses for “advanced integrated circuits”, and any products that use them, if a company cannot certify that no U.S. entity is seeking those chips and that exporting them will not disrupt domestic supply. This mechanism is designed to ensure that U.S.-produced technologies meet the domestic demand of U.S.-based customers, including small businesses and academic institutions, before chipmakers satisfy international demand.

The GAIN AI Act brings the same tradeoffs as past proposals to manage the global diffusion of advanced AI chips, however. It highlights the problem of how the government should best intervene – given the dual-use nature of AI technology, with both civilian and military uses – without undermining long-term U.S. AI leadership and competitiveness in international markets. Consequently, policymakers must navigate this challenging environment in a way that avoids regulatory overreach that could undermine the very innovation it seeks to protect.

Chip Market Structure and Access

Global Dominance

The semiconductor industry is expected to grow into a trillion-dollar market by the end of the decade. This growth relies on chips that are critical across global industries, powering everything from computers and smartphones to advanced AI systems. The United States currently holds a leading global position in the architecture and design of advanced semiconductors essential for AI workloads. U.S. firms such as NVIDIA and AMD have successfully diversified their global sales, though the United States remains their most critical market. In 2024, neither company derived more than 50 percent of its revenue from a single country: NVIDIA’s U.S. revenue share stood at 47 percent, while AMD’s was 34 percent. Notably, while China represents AMD’s second-largest market at 24 percent of revenue, it accounts for only 13 percent for NVIDIA, placing it behind Singapore and Taiwan. This global structure underscores U.S. technological leadership while highlighting the delicate balance between domestic control and international market reliance.

The Structural Vulnerability to Chip Shortage

Adding to the challenge of this global access distribution is the structural fragility of the semiconductor industry. The sector is fundamentally susceptible to production cycles due to the requirement for extremely long lead times, billions in upfront capital, and volatile, innovation-driven demand. The existing global chip shortage has significantly impacted markets, and a future crisis is highly probable. An analysis by Bain and Company projects that the surge in AI-driven demand could increase the need for certain components by an additional 30 percent by 2026. Coupled with the anticipated acceleration of smartphone and PC replacement cycles by AI-enabled devices, the ongoing geopolitical tensions could generate the next semiconductor shortage.

Domestic Concentration and Access Crisis

In the United States, it’s becoming easier for people to access AI tools and models, but the chips and computing power that makes these systems run is far from accessible. Many depend on a small group of cloud hyperscalers – large cloud service providers that operate massive-scale data centers that provide infrastructure for AI and other workloads. These hyperscalers can afford to build and maintain a massive infrastructure containing thousands of AI chips; they then rent AI capacity to smaller companies, such as startups and universities. Most of the advanced chips produced by U.S. manufacturers end up in the hands of these giants. As a result, computational power is concentrated in just a few firms that have the scale and capital to support it. Thus, for U.S. startups and medium companies, the problem is not fully about chip shortages; it’s that they can’t realistically own it. High costs, opaque pricing, and long waits for hardware mean that these companies are effectively incapable of buying the chips.

The current market structure is complex. The United States is a global leader in advanced semiconductor technology, yet the market is exposed to major supply risks. These realities provide the essential context for evaluating policy interventions, such as the GAIN AI Act, which aims to prioritize domestic users but could also disrupt the global markets that sustain U.S. chipmakers.

The Unintended Consequences of the GAIN AI Act

While the GAIN AI Act aims to address domestic supply constraints for startups and universities, data show that the issue for these companies is not that the chips do not exist but that they can’t acquire them. Yet the legislation risks harming long-term U.S. competitiveness and innovation capacity of chips and AI developments. By mandating the satisfaction of domestic quotas before fulfilling international sales, both domestic and international customers face extra licensing paperwork and compliance costs in a market that has already long waiting times. Further, these controls risk the redistribution of market share to foreign competitors operating without similar restrictions. If U.S. firms retreat from international markets, other firms will fill the gap; estimates show that South Korean firms could gain $21 billion in sales, European Union firms $15 billion, and Taiwanese firms $14 billion of U.S. firms’ losses.

In fact, the legislation may effectively deny U.S. chipmakers the opportunity to maximize sales in global markets, even when they have successfully competed and established customer relationships. The legislation may threaten U.S. innovation since the highly specialized semiconductor industry relies on maximizing global revenue to cover the billions required for the next generation of chip research and development.

Critics also argue that the GAIN AI Act goes further than past controls, noting that the legislation would apply even to older AI chips which were specifically developed to comply with prior export controls, such as Nvidia’s HGX G20. The scope of technologies subject to this bill is also not limited to those posing national security concerns, potentially expanding controls to chips not currently requiring a license, such as those used in gaming consoles. In sum, the legislation would unintentionally raise compliance costs for U.S. companies and add regulatory uncertainty to an already volatile semiconductor market.

Conclusion

Evaluating proposals to address AI chip export controls, including the GAIN AI Act, ultimately requires policymakers to define the fundamental objective of U.S. dual-use technology policy. Given the United States’ leadership in advanced chip creation, the path forward presents two distinct approaches. Policymakers will either adopt a narrow, clearly targeted export policy rooted in concrete security concerns – keeping U.S. technology competitive at home and abroad – or impose broad restrictions on global sales that could unintentionally weaken the country’s market position.

The structure of the chip market and its supply chain suggests that the most effective strategy for protecting U.S. advantage in AI is accelerated domestic capacity and innovation. That means the federal government should move away from limiting the market to empowering U.S. companies to lead and shape global standards. This calls for a more focused approach that keeps export restrictions narrow and grounded in real security risks, while reducing barriers to speed up domestic production of both chip manufacturing and AI infrastructure.