In the high-stakes world of artificial intelligence, NVIDIA's first-ever nuclear investment speaks volumes about an industry facing an existential threat. The GPU giant's participation in TerraPower's $650 million August funding round—alongside Bill Gates and HD Hyundai—isn't just another tech diversification play. It's a calculated bet on nuclear energy as the solution to AI's most pressing constraint: power.
But for investors scrutinizing TerraPower's impressive $3.4 billion in total funding against actual technical progress, the numbers reveal a concerning gap. While the company has secured $160 million from DOE and access to a $3.2 billion program, these figures stand in stark contrast to the fundamental HALEU supply bottleneck that threatens the entire commercialization timeline.
The collision between AI's exponential growth and the grid's linear expansion is creating what NVIDIA CEO Jensen Huang bluntly described as requiring "redo[ing] the entire grid." Data centers already consume 2-3% of U.S. electricity, projected to reach 9% by 2030. AI inference power demand is growing at a staggering 122% compound annual rate through 2028. Meanwhile, NVIDIA's upcoming GPU server racks are expected to draw significantly more power than previous generations—potentially enough to power hundreds of homes.
"Nuclear energy is going to become a more vital energy source," stated Mohamed Siddeek, NVentures corporate vice president, in what might be the understatement of the year. The question isn't whether AI needs nuclear power—it's whether TerraPower can deliver it in time.
AI Power Crisis Drives Investment Logic
Goldman Sachs forecasts a 165% increase in global power demand from data centers by 2030 compared to 2023 levels. Current power usage by data centers sits at approximately 55 gigawatts, with AI workloads accounting for 14%—a figure set to multiply rapidly.
This isn't a distant problem. A U.S. Department of Energy commissioned report projects data centers will consume between 6.7% and 12% of total U.S. electricity by 2028—equivalent to 325 to 580 terawatt-hours. As Jean-Marie Brunet of Siemens EDA noted: "Power is not a joke anymore."
TerraPower's strategic partnership with Sabey Data Centers reveals the company's recognition of this opportunity. Their memorandum of understanding focuses on integrating Natrium reactors directly with new data center builds in the Rocky Mountain region and Texas—areas already straining under data center growth. Yet a close examination of this partnership reveals it as merely an agreement to explore various project execution structures, not a binding financial commitment that would validate TerraPower's commercial timeline.
Natrium's Technical Promise Meets Supply Chain Reality
TerraPower's Natrium reactor design offers genuine innovations that address AI's specific power needs. The 345-megawatt sodium-cooled fast reactor incorporates a molten salt energy storage system that allows for peak output of 500 megawatts during high demand—precisely when data centers need it most.
The reactor operates at temperatures exceeding 350°C, utilizing passive cooling mechanisms that enhance safety while reducing costs. A key innovation is the separation of energy and nuclear systems, enabling non-nuclear teams to manage operations outside the nuclear control area—a design choice that improves both safety and operational efficiency.
The Department of Energy's positive environmental impact assessment for a critical experiment related to the Natrium reactor provides external validation of the approach, though this represents just one step in a lengthy approval process. The project is part of the DOE's Advanced Reactor Demonstration Program, with a total investment of up to $2 billion matched by TerraPower and partners.
But here's where the glossy marketing collides with reality: TerraPower's project faces a minimum two-year delay due to challenges in securing high-assay low-enriched uranium (HALEU) fuel. This isn't a minor supply chain hiccup—it's a fundamental obstacle that undermines the company's timeline claims and raises questions about whether TerraPower can deliver power when NVIDIA and other tech companies will need it most.
HALEU Supply Chain Exposes the $650M Question
The HALEU fuel challenge exposes how TerraPower's ambitious investor presentations gloss over a fundamental supply chain reality that threatens not just their timeline but the entire advanced nuclear sector's credibility. The numbers reveal a disconnect between marketing and manufacturing capacity that investors should find alarming.
Following Russia's invasion of Ukraine, the primary commercial source of HALEU became politically untenable, forcing TerraPower to develop alternative supply chains from scratch. The Department of Energy projects domestic HALEU demand could reach 50 metric tons annually by 2035. Currently, only about 700 kilograms have been produced domestically since 2023—less than 2% of the projected annual need.
TerraPower is pursuing multiple parallel strategies to address this gap. The company executed a term sheet with ASP Isotopes Inc. to expand global HALEU production, including plans for an enrichment facility in South Africa with a 10-year supply agreement. Domestically, TerraPower has agreements with Centrus Energy to utilize its NRC-licensed HALEU production facility.
Additionally, TerraPower partnered with Oklo and Hexium to develop domestic production using Atomic Vapor Laser Isotope Separation (AVLIS) technology. The DOE has made conditional commitments to provide HALEU to TerraPower and awarded contracts to six companies to enhance the U.S. supply chain, with each receiving a minimum of $2 million and potential for $800 million over ten years.
Despite these efforts, the fundamental question remains: Can TerraPower secure sufficient HALEU to meet its revised timeline, or will further delays push commercial operation beyond 2031?
Leadership Credentials vs Execution Challenges
TerraPower's leadership team brings impressive credentials to these challenges. CEO Chris Levesque has over 30 years of nuclear experience, including service in the U.S. Nuclear Navy. Bill Gates serves as Chairman, providing both financial backing and strategic vision. The recent addition of Thomas Kuhn, former President and CEO of the Edison Electric Institute, brings over 30 years of utility industry experience to the board.
In May 2025, TerraPower expanded its executive team with Eric Williams as Chief Operating Officer, who brings 22 years of engineering experience focused on nuclear power plant design and safety. Steven Hellman joined as Chief Financial Officer, contributing 35 years of experience in the global energy sector, particularly in mergers and acquisitions.
This team has made impressive regulatory progress. A close examination of TerraPower's March 2024 construction permit application—the first for a commercial advanced reactor—reveals strategic timing that allows the company to claim regulatory progress while obscuring the fundamental HALEU supply challenge that remains unresolved in its investor materials.
The NRC—an agency with a history of regulatory capture by the industry it oversees—claims to be ahead of schedule on its safety review, projecting completion by June 2026, conveniently aligning with TerraPower's revised timeline. The commission has already completed a draft safety evaluation ahead of schedule and issued an Environmental Assessment and Finding of No Significant Impact for TerraPower's exemption request.
Construction of the non-nuclear portion of the plant began in June 2024, with nuclear construction expected to start in 2026 pending NRC approval. Commercial operation is projected for 2031—a timeline that assumes resolution of the HALEU supply challenges.
Investment Analysis Shows Funding vs Fundamentals Gap
Private investors have committed $1.4 billion to TerraPower, with the company's total funding exceeding $3.4 billion when including government support—a substantial sum that still may not cover additional costs if HALEU-related delays continue. The company has secured $160 million from the Department of Energy for advanced reactor development, with access to the DOE's $3.2 billion Advanced Reactor Demonstration Program.
In comparison, X-energy's Xe-100—an 80MWe reactor scaling to 320MWe—is targeting operations by 2029. Both TerraPower and X-energy are considered better positioned than NuScale, which has faced challenges with its first-of-a-kind small modular reactor project.
TerraPower's international ambitions are evident in its initiation of the UK regulatory approval process. However, the partnership with Sabey Data Centers—while strategically positioned—represents a non-binding memorandum of understanding rather than the concrete commercial commitment that would validate the company's timeline claims.
The Nuclear-AI Convergence Gamble
The convergence of AI's energy crisis and TerraPower's advanced nuclear technology creates a compelling investment case despite significant execution risks. For philanthropist investors, TerraPower represents a rare opportunity to back a nuclear startup with genuine technical innovations, strong government support, and strategic industry partnerships that address both climate goals and AI infrastructure needs.
However, the HALEU fuel supply challenge remains a critical risk factor that could further delay the project beyond current projections. The gap between TerraPower's impressive funding rounds and its ability to deliver on timeline promises should concern investors who've seen too many sophisticated marketing narratives obscure fundamental operational challenges.
NVIDIA's investment signals that tech giants are no longer content to wait for energy solutions—they're actively backing the technologies they'll need to power AI's future. The question for investors is whether to follow NVIDIA's lead or wait for TerraPower to demonstrate actual progress on domestic HALEU production before committing capital to nuclear's AI-powered renaissance.
The next 12-18 months will be decisive as TerraPower navigates final regulatory approvals and proves it can bridge the gap between ambitious projections and supply chain realities. For investors, the choice is clear: bet on TerraPower's ability to execute against the odds, or recognize that even the most sophisticated nuclear technology means nothing without the fuel to run it.
Things to follow up on...
-
NRC regulatory capture: The Nuclear Regulatory Commission has faced allegations of being captured by the nuclear industry it regulates, particularly after the Fukushima disaster, which raises questions about oversight effectiveness for new reactor approvals.
-
X-energy competitive positioning: X-energy's Xe-100 reactor offers an 80MWe design scaling to 320MWe with operations targeted by 2029, making it a direct competitor to TerraPower's timeline in the advanced reactor market.
-
DOE HALEU contracts: The Department of Energy awarded contracts to six companies including BWXT, Centrus, Framatome, GE Vernova, Orano, and Westinghouse to enhance U.S. HALEU supply chains, with each receiving minimum $2 million contracts and potential for $800 million over ten years.
-
TerraPower UK expansion: TerraPower has initiated the UK regulatory approval process in partnership with KBR, building on its U.S. regulatory achievements to pursue international market opportunities beyond the domestic market.