Recently focusing on the energy investment directions for 2026, I discovered an interesting phenomenon—the story of new energy has completely changed.

In the past few years, everyone was talking about electric vehicle subsidies and solar overcapacity; now the focus has shifted directly to AI's rigid demand for electricity. According to the latest forecasts from IEA and Goldman Sachs, global data center electricity consumption will surge from 460 TWh in 2022 to about 1,050 TWh in 2026, with AI-related parts contributing over half of the growth. The power consumption of training a large AI model is equivalent to the annual electricity usage of tens of thousands of households.

What does this mean? Traditional solar and wind power cannot meet the 24/7 stable power supply requirements of AI data centers. Tech giants like Microsoft, Amazon, and Google are now heavily investing in nuclear energy. Amazon plans to deploy 12 small modular nuclear reactors, and Google has committed to tripling nuclear capacity by 2030. This is not hype but real capital flow.

But I notice a more critical bottleneck—power generation is easy, but transmission is difficult. The global power grid is severely aging, with delivery times for high-voltage transformers and switchgear reaching 2-3 years, and supply shortages are expected to continue until at least 2027. This directly drives the revenue growth rate of power companies from 1% to 4-6%.

In Taiwan, green energy concept stocks like Delta Electronics and Huasheng Electric are benefiting from grid upgrades and the high power density demand of AI servers. Delta's orders exploded in 2025, and Huasheng is a long-term partner of Taipower and a leader in Taiwan’s electric vehicle charging station industry. Traditional green energy stocks like United Renewable Energy, Shangwei, and Yuanjing, though less volatile, are also maintaining steady growth under the long-term trend of global net-zero emissions transformation.

In the US stock market, Constellation Energy, the largest nuclear operator in the US, has signed a 20-year restart contract for Three Mile Island with Microsoft, with a major expansion of data center projects expected in 2026. Eaton, a leader in smart grid technology, now faces transformer demand with a 24-month lead time. GE Vernova benefits from global grid upgrade investments, with order backlogs reaching new highs. NextEra Energy, the largest renewable energy company in the US, offers a defensive allocation amid the net-zero transition.

The investment logic of green energy concept stocks has shifted from policy subsidies to demand-driven. It is recommended that AI electricity stocks constitute 50-60% of the portfolio as growth drivers, traditional energy stocks account for 30-40% as a defensive base, and the remaining 10% in cash or bonds as a buffer. The key is to monitor leading indicators such as AI capital expenditure, grid investment scale, and order backlog, rather than chasing hype.

Between 2026 and 2030, this time window, new energy stocks are indeed the most promising structural opportunities to deepen.