A top-performing sustainable investment fund has found its answer to AI's power crisis not in solar panels or wind turbines, but in Japan's existing nuclear reactors and industrial infrastructure. The BNP Paribas Green Tigers Fund, managed by Oscar Yang at Impax Asset Management, has returned 34 percent year-to-date by concentrating on Japanese nuclear, grid, and industrial energy efficiency stocks, making it Asia's best-performing dark green fund.

The $208 million portfolio carries the European Union's highest sustainability classification, yet its composition looks nothing like traditional green investing. Rather than betting on renewable energy developers, Yang has steered capital toward companies positioned at the intersection of nuclear infrastructure, industrial grid hardware, and energy efficiency services that artificial intelligence (AI) builders increasingly need but cannot easily build themselves.

Why Is AI Creating Such an Urgent Power Problem?

The numbers behind this shift are stark. Global data center electricity consumption is on track to hit roughly 1,100 terawatt-hours (TWh) in 2026, which is approximately double the 460 TWh consumed in 2024 and equivalent to Japan's entire national electricity output. In the United States alone, data center energy demand is forecast to nearly double from 80 gigawatts (GW) in 2025 to 150 GW by 2028, driven almost entirely by AI training and inference workloads.

The supply-demand mismatch is becoming critical. New generation capacity is coming online at only 2 to 3 GW per year while data centers are adding 5 to 7 GW of new demand annually. Nearly half of planned U.S. AI data centers for 2026 are already delayed, and New York has moved to ban new hyperscale facilities outright. The traditional green energy playbook of adding more solar and wind capacity does not solve a problem that is fundamentally about dispatchable baseload power, meaning electricity that can be reliably delivered on demand around the clock.

What Makes Japan's Position Unique in Solving This Crisis?

Japan has something neither the United States nor Europe can quickly replicate: a fleet of nuclear reactors that already exist, a heavy industrial base capable of servicing them, and a government now actively accelerating their return to service. As of early 2026, 14 reactors have been restarted under post-Fukushima safety standards. In January, Tokyo Electric Power Company (TEPCO) restarted Unit 6 of the Kashiwazaki-Kariwa Nuclear Power Station after 14 years offline, following approval from Niigata Governor in November 2025.

Japan's government is leaning into this moment with concrete policy support. The government's target is nuclear providing 20 to 22 percent of electricity by 2030, a policy reversal driven explicitly by energy security and data center demand. Additionally, a $1.34 billion clean power subsidy program covering up to 50 percent of capital expenditure for qualifying projects, including data centers, started in fiscal 2026. The requirement is strict: 100 percent of electricity must come from renewable or nuclear sources. That condition alone makes Japan's nuclear-industrial complex a structural beneficiary, not a cyclical one.

How to Identify Companies Positioned to Benefit from AI's Power Demands

• Nuclear Infrastructure Operators: Companies with existing reactor fleets and engineering expertise to restart and maintain nuclear facilities, which provide the dispatchable baseload power that AI data centers require around the clock.
• Industrial Grid Hardware Providers: Manufacturers of power transmission equipment, transformers, and grid modernization technology that connect nuclear plants to data centers and ensure reliable electricity delivery.
• Energy Efficiency and Digital Platform Companies: Firms offering industrial optimization software and digital platforms that help manage power consumption across manufacturing, transportation, and facility networks.

Hitachi sits at the center of Yang's investment case. The company straddles every relevant layer: modern power grids, nuclear plant engineering, and physical AI infrastructure, with partnerships in place with Google Cloud, OpenAI, Anthropic, and Intel that tie its Lumada digital platform to transformer networks, rail systems, and industrial facilities. Its earnings grew 30.3 percent over the past year.

Mitsubishi Heavy Industries offers exposure through a different angle: a record order backlog of ¥10.77 trillion and surging demand for gas turbine, nuclear, and carbon capture projects. Marubeni spans from nuclear to liquefied natural gas (LNG) to grid infrastructure. These are not AI hype plays. They are industrial companies with decades of operational history that happen to be positioned exactly where the energy bottleneck is most acute.

The Green Tigers performance should prompt a harder look at how "dark green" investing has been practiced versus how it is now being tested. For years, European Union Article 9 funds defaulted to the same shortlist: solar developers, wind turbine makers, electric vehicle battery suppliers. Yang's fund holds that classification and has almost nothing in common with that playbook. The 34 percent return, while markets in much of the rest of Asia have struggled, is a verdict on the approach.

The AI energy crisis is reshaping what responsible investing actually means in practice. Capital that wants to be green and wants to be useful, to actually facilitate the decarbonization of an electricity grid strained by AI demand, has to follow the problem, not the aesthetic. The problem right now is dispatchable, low-carbon baseload power. Japan has it. The United States and Europe are scrambling to build it.

Microsoft's $1.6 billion deal to restart Three Mile Island's Unit 1 via Constellation Energy, with the 835 megawatt (MW) facility expected online by 2027, is the clearest American acknowledgment that the wind-and-solar answer does not work alone at this scale. Japan figured this out faster, and its industrial names are already booking the orders that prove it.

Yang's $208 million fund is a small number against the broader green capital universe. But it is the right number to watch. It tells you where the energy transition is actually going, as opposed to where it was assumed to be going three years ago.