Author:BITWU.ETH
Recently, I followed Professor Chen Xiaodong's NTU course "Social Impact of Disruptive Technologies" and gained a new understanding of nuclear power.
I have always thought nuclear power is a typical "national-level project": requiring massive investment, long cycles, and complex approvals, making it difficult for ordinary enterprises and capital to truly participate.
But the emergence of AI data centers might be changing this logic.
The biggest change is not that nuclear technology has suddenly matured, but that nuclear energy has, for the first time, encountered commercial buyers who are sufficiently powerful and urgently in need.
In the past, nuclear power primarily addressed national energy security concerns; now, AI companies like Microsoft, Google, and Amazon are also actively seeking energy solutions that can provide long-term, stable, low-carbon electricity.
As a result, several changes are happening simultaneously:
Large-scale nuclear power still belongs to national infrastructure; small modular reactors have begun to enter commercial validation and are more suitable for the power scale of data centers, industrial parks, and high-energy-consumption enterprises.
At the same time, passive safety technology, modular manufacturing, and improvements in regulatory efficiency are also attempting to gradually transform nuclear energy from a "one-time mega-project" into a replicable industrial product.
The value of nuclear energy is not just electricity generation.
Heating, industrial steam, hydrogen production, seawater desalination, and even stable energy supply for high-energy-consumption manufacturing could become new commercial scenarios.
What this corresponds to is not just the nuclear power plant itself, but also an entire supply chain including fuel, forgings, instrumentation and control, high-temperature superconducting magnets, and more.
Of course, nuclear energy is not without risks.
Construction cycles, cost overruns, nuclear waste, fuel supply, regulation, and project financing—any problem in these areas could turn a project from a long-term asset into a stranded asset; fusion is still far from true commercialization.
So my current assessment of nuclear energy is:
AI has not made nuclear energy mature overnight, but it is providing nuclear energy with commercial demand, capital support, and real-world orders it never had before.
In the next decade, nuclear energy may not be the sexiest narrative in the AI industry, but it could become one of the most underestimated pieces of infrastructure.
