Anthropic Energy Analysis Report

For the United States to Lead in AI: The Imperative of Infrastructure Investment

As the landscape of technology continues to evolve, artificial intelligence (AI) stands at the forefront, reshaping our economy and society. To maintain its competitive edge in this critical sector, the United States must significantly invest in the underlying resources that enable AI development. This involves not only ramping up computing power but also ensuring a reliable energy supply. At Anthropic, we recognize the daily impact of AI models across various sectors, from businesses to government institutions. The time to act decisively is now; together, we can ensure that the AI future is built in America.

The Energy Demands of Frontier AI

One of the primary challenges facing AI development is its voracious appetite for energy. Training frontier AI models requires continuous access to firm, reliable power sources. With forecasts suggesting a dramatic increase in energy needs—projecting anywhere from 20 to 25 gigawatts by 2028 just for AI training—the urgency to bolster the United States’ energy capacity cannot be overstated. For context, this demand is comparable to double that of New York City’s peak electricity usage.

Comparatively, China is actively ramping up its energy infrastructure, adding over 400 gigawatts of power capacity last year alone. If the U.S. is to compete, a strategic, “all-of-the-above” approach to energy sources is essential. This encompasses traditional energy, alongside emerging technologies like advanced nuclear and next-generation geothermal energy. By developing these resources efficiently, we can better meet both immediate energy needs and long-term AI infrastructure requirements.

Policy Recommendations for Infrastructure Development

In light of these pressing requirements, we’ve formulated a set of policy recommendations aimed at building the necessary infrastructure for AI development in the U.S. The roadmap comprises two essential pillars: large-scale AI training facilities, and broader infrastructure for nationwide AI deployment.

Pillar 1: Large-Scale AI Training Infrastructure

Developing large-scale training facilities is critical for advancing American AI capabilities. Recommendations include:

1. Utilizing Federal Lands: Make federal lands available for constructing AI infrastructure, circumventing lengthy state and local zoning processes.

2. Accelerating NEPA Processes: Expedite the National Environmental Policy Act (NEPA) review processes for AI infrastructure projects, including conducting advance reviews.

3. Public-Private Partnerships: Encourage collaborations with private sectors to fast-track power line infrastructure upgrades and buildout.

4. Improving Interconnection Processes: Support utilities in reforming key interconnection processes, potentially employing AI to enhance efficiency. In cases vital to national security, leverage federal authorities to prioritize these connections for AI training.

Pillar 2: Broad-Based Infrastructure for AI Innovation

To ensure the widespread deployment of AI technologies, broader infrastructure changes are required:

1. Energy Permit Acceleration: Streamline permitting processes for geothermal, natural gas, and nuclear energy infrastructure.

2. Establishment of Electric Transmission Corridors: Set up National Interest Electric Transmission Corridors to hasten the permitting of transmission lines in key areas of AI data center growth.

3. Strengthening Domestic Production: Foster domestic production of essential grid components and turbines via loan programs and strategic reserves.

4. Workforce Development: Develop training, apprenticeship, and entrepreneurship programs targeted at critical energy and construction workers, ensuring a skilled workforce can meet growing energy demands.

The Road Ahead: Addressing Regulatory Challenges

While steps have already been taken to reduce barriers to energy development, such as ambitious nuclear power targets, comprehensive action is necessary. Addressing regulatory hurdles that delay energy project timelines is crucial if the U.S. is to catch up with international competitors.

Policymakers can expedite the buildout of large-scale infrastructure for AI training in strategic locations by:

• Supporting initiatives that expedite permitting.
• Promoting collaborative efforts with utilities to improve grid interconnection.
• Streamlining reviews through established channels like the U.S. Army Corps of Engineers to speed up Clean Water Act assessments for AI data centers.

The Potential for AI to Drive Energy Innovation

The intersection between AI and energy innovation holds immense promise. With an evolving energy landscape, AI has much to offer, from increasing the efficiency of energy production to optimizing consumption patterns. By embracing next-generation technologies, the U.S. can solidify its role as a global leader in AI development while simultaneously driving advancements in energy.

Conclusion: A Call to Action

America possesses the economic strength, technological expertise, and innovative spirit to overcome these infrastructure challenges. By implementing the recommendations outlined in our report “Build AI in America,” we can effectively harness existing federal authorities to dismantle the regulatory barriers that have historically impeded energy development. Through collaboration among government, industry, and communities, a robust foundation for decades of American leadership in AI can be established.