From Minerals to Memes

US and EU policymakers have identified artificial intelligence (AI) as a key driver of competitiveness and a major societal disruptor. Recent initiatives such as the EU’s AI Act, the Competitiveness Compass, and the AI Continent Action Plan, seek to improve Europe’s position in AI development and deployment. Across the Atlantic, the Trump administration’s AI Action Plan includes policy recommendations to promote AI across three pillars: innovation, infrastructure, and national security. These AI initiatives share certain similarities, including a push to simplify or eliminate regulation and expand AI infrastructure. Both sides, therefore, are prioritizing innovation and adoption over guardrails; the word “safety” appears once in the US AI Action Plan and twice in the EU AI Continent Action Plan. Still, early signs of tension in international AI policy approaches emerged after the February AI Action summit in Paris, whose final declaration the United States and the United Kingdom opted not to sign. And unilateral trends, such as the EU’s emphasis on digital sovereignty and the US push for export dominance via the Executive Order on Promoting the Export of the American AI Technology Stack, may complicate transatlantic coordination on AI.

GMF Technology’s Transatlantic Tech Exchange strives to strengthen transatlantic cooperation on technology policy and uncover overlooked areas for collaboration in the face of such tensions. This year’s iteration is centered on the concept of “The AI Value Chain: the natural resources, industrial and network infrastructure, data, and public and private investments and inputs that create the machine learning systems known or marketed as AI”. Over the next year, GMF Technology will explore this concept to:

• build trust and mutual understanding among European and US lawmakers working on AI at the federal and state levels
• discover areas of possible cross-party and transatlantic policy alignment
• deepen understanding of EU and US technology policy and innovation environments
• offer recommendations to advance innovation and democratic values in technology innovation and policy

By examining the AI value chain, policymakers can broaden their perspectives on fostering innovation in a less siloed and more holistic manner, and steer technological development and governance in a direction that advances the public interest and creates democracy-affirming technology pipelines. The effort aims to bring structure to AI innovation and policy conversations, identify spillover effects, and unearth overlooked areas of transatlantic coordination.

The concept of an AI value chain, or AI stack, is already present in policy discussions. The EU’s AI Act outlines “Responsibilities along the AI Value Chain”, and legal interpretations of the act define the AI value chain as the suite of roles tied to an AI product, alongside compliance obligations. The US AI Exports executive order seeks to advance an “American AI technology stack”. Researchers have defined the AI value chain as “the organizational process through which an individual AI system is developed and then put into use (or deployed).” GMF Technology’s work on the AI value chain builds on the team’s “technology stack” framework, which examines penetration of a country’s technology and governance ecosystem by adversarial actors.

The value chain maps AI from start to finish, and presents the overlapping policy and innovation issues across the chain. They are:

• infrastructure
  • natural resources: energy, water, and minerals
  • semiconductor fabs, data centers, and broadband
• building blocks: semiconductors, data, and cloud services
• algorithmic and machine learning models and systems
• integration: consumer-facing applications, enterprise or B2B software, industrial applications

The diagram illustrates the cross-cutting policy issues that span these nodes, and which intersect from local to international levels. Technology innovation and policy efforts often target individual pieces of the value chain using different tools: Countries pursue critical mineral partnerships, invest in manufacturing through CHIPS acts such as those passed by the EU and the United States, and weigh the benefits and costs of a growing data center industry. Initiatives to spur AI innovation and harness local competitive advantages, such as the Massachusetts AI Hub, are launching, with dedicated funds aimed at scaling up compute capacity and supporting startups. Recognizing opportunities across the entire value chain will bolster US and EU leadership in AI, yet the nodes are often addressed only disjointedly. A specialist in privacy legislation, for example, may not consider industrial policy for minerals, data centers, or semiconductors. At the same time, data retention, data minimization, governance, and copyright policies all have implications for demand projections that shape the scale of data center or semiconductor fab buildouts, and their energy and resource needs. A value chain framing can draw connections and highlight unexpected policy spillovers across the nodes of the chain.

GMF Technology aims to offer this framing while addressing questions along the AI value chain that have important implications for the transatlantic relationship. Key policy questions include:

Infrastructure

• How will the rising resource demands of physical infrastructure, especially data centers, incentivize efforts to develop less resource-intensive AI models?
• Will mitigating and responding to AI-related vulnerabilities in critical infrastructure present an opportunity for cross-border and public-private collaboration?
• How will the EU and the United States update their grids to manage new AI-related demand, including through the use of new grid management technologies?
• How can the EU and the United States harness comparative advantages and achieve the shared goal of secure semiconductor supply chains in a time of geopolitical and trade tensions?

Building Blocks

• The resources required to develop and deploy AI are highly concentrated. How can lawmakers spur innovation by expanding access to computing power and data?
• Export controls are a key element in ensuring control of the AI value chain from adversaries. With key nodes on both sides of the Atlantic, how will the EU and the United States continue to cooperate?
• Will a patchwork of data privacy laws across US states and efforts to simplify the EU’s General Data Protection Regulation alter the data governance landscape?
• How can the EU’s and the United States’ cloud initiatives advance competitiveness? Are there opportunities for joint scientific and defense research?

Algorithms and Models

• The EU and the United States have leaned toward open-source strategies to, respectively, decrease market concentration and counter adversaries such as China. Where can further collaboration on open-source or open-weight models occur?
• Where can the EU and the United States find common ground on agentic AI, including competition policy, cybersecurity challenges, privacy, and legal and compliance issues?
• Could narrow cases such as public data sets be an area of transatlantic cooperation on AI innovation and applications that benefits the public interest?
• How will areas of shared concern such as biosecurity facilitate data sharing initiatives?
• Will transatlantic trade tensions linked to EU digital laws lead to retrenched enforcement of or fewer requirements for testing AI systems for reliability, accuracy, and safety?

Integration

• How will the US approach cooperation with partners such as the EU to counter China in multilateral bodies, as outlined in the Trump administration’s Action Plan?
• How can the EU and the United States mutually develop AI-ready workforces, mitigate job loss, and improve the AI talent pipeline?
• Will states lead the way on risk management of AI deployment and applications, which could affect hiring or housing, among other areas? How will they balance guardrails with the US AI Action Plan’s and the EU AI Continent Action Plan’s emphasis on AI adoption?
• Many firms straddle the Atlantic, with operations in the EU and the United States. Can the inherent transatlantic dimension of many AI industrial applications, and their supply chains, facilitate transatlantic coordination?

Read more about the Transatlantic Tech Exchange here.