Nuclear, AI, and Data Centers Take the Spotlight at ARPA-E Energy Summit 2025

In March 2025, with support from a Kleinman Center Student Grant, I participated in ARPA-E’s Student Program and attended the ARPA-E Energy Innovation Summit in Washington, D.C. The Department of Energy’s ARPA-E agency funds high-impact applied R&D to turn scientific discoveries into breakthrough energy technologies, and the annual summit brings together scientists, entrepreneurs, investors, and policymakers to advance the future of energy. 

Amid excitement around inflection points in both nuclear fusion and advanced fission, a consistent throughline was the need to accelerate deployment. One of the standout panels advocated for collaboration between fusion and fission, which have historically been siloed. Experts discussed how a unified approach to supply chains, workforce needs, regulatory frameworks, and manufacturing could collectively speed the deployment of modular fission designs and bring fusion companies closer to net energy gain.

“When people ask—why should I be excited about fusion now? —it’s why you should have been excited about computers when the transistor was invented.” – Phil Larochelle, Partner at Breakthrough Energy Ventures

There’s good cause for excitement. Fusion is entering its “transistor era” —component level breakthroughs like Commonwealth Fusion System’s superconducting magnets, cheaper solid-state lasers, pulsed power advancements, and rapid simulation tools are accelerating fusion’s technology readiness level. On the fission side, advanced reactor designs are theoretically cheaper to construct and open up new market segments (e.g. industrial heat, data centers, remote military bases). These new reactor designs are starting to pass regulatory approval – Westinghouse’s ARPA-E-funded eVinci microreactor recently announced an Nuclear Regulatory Commission (NRC) licensing milestone. 

At the same time, there are critical financing, policy, and workforce barriers facing both advanced fission and fusion. Despite movement on permitting reform and potential ADVANCE Act 2.0 legislation, the NRC licensing process for new advanced fission reactor designs is still highly complex and can cost in the ballpark of $50-$100M. Plus, customers hesitate to be first to finance expensive first-of-a-kind reactor builds. Fusion is chronically underfunded in the U.S., with China outspending the U.S. government 2-to-1. While fusion avoids NRC hurdles, it lacks a regulatory framework for issues like tritium management and remote maintenance. 

From a policy perspective, panelists called for expanding successful public-private models like NASA’s space exploration programs, the Advanced Reactor Demonstration Program (ARDP), and Project Pele, which de-risk early development and deployment through milestone-based cost-sharing. A fusion equivalent of ARDP, tied to technical milestones, could unlock private capital. Regulatory streamlining of licensing of first-of-a-kind models through the Department of Defense (DOD) and Department of Energy (DOE)’s licensing authority could accelerate timelines and lower costs. Both technologies would also benefit from modernized grid interconnection policies that enable seamless integration with emerging high-load users like data centers. Finally, panelists called for removing statutory barriers to cross-agency collaboration (DOE, DOD, NRC) on fission and fusion R&D and for federal investment in workforce development programs that link universities, national labs, and private industry to build nuclear talent pipelines. 

AI-driven data center growth was everywhere. The Technology Showcase floor was packed with advanced cooling and battery technologies for data centers. ARPA-E Fellows pitched R&D in baseload power generation (fusion, fission, geothermal), hardware for accelerated computing, and long-term energy storage. Across all events, it was clear that rising electricity demand and data center growth are driving energy innovation. 

A panel with NVIDIA and National Grid Ventures discussed how AI-driven load growth is triggering a fundamental shift in utilities’ decades-long mindset of building “just enough.” Yet, while data centers can be built in months, transmission takes a decade. NVIDIA emphasized that today’s data centers are starting to act more like grid assets—co-locating with generation, reusing heat, and offering flexibility. The U.S. now needs effective policy tools to integrate demand-side resources and flexible loads at scale. Panelists highlighted behind-the-meter generation and grid-edge intelligence as promising but untested paths forward, while also warning that failing to deploy innovative resources now could leave the grid unprepared for the surge in AI demand.

Despite an undercurrent of regulatory uncertainty, it was impossible not to leave the summit excited about energy innovation —from advanced nuclear microreactors to bioengineered crops that extract critical minerals to quantum simulations that accelerate materials discovery. To meet rising energy demand, we need matching innovation in financing models, public-private partnerships, regulatory pathways, workforce development, and commercialization strategies to deploy these breakthroughs at scale.

Olga Belyanina

MBA Student, Wharton School

Olga Belyanina is an 2025 MBA Candidate at The Wharton School studying Entrepreneurship & Innovation and Business, Energy, Environment, and Sustainability.