這篇以 SMR 與 AI 資料中心的插圖作為主視覺，對應的是一個越來越清楚的問題：當算力需求上升，能源供給是否能跟得上？
Chinese
如果我們接受一個前提：AI 的基礎是算力，那麼下一個問題就很直接， 算力的基礎是什麼？答案不是模型，而是能源。沒有足夠、穩定、可長期擴張的電力， 所有關於大模型、推論服務、邊緣 AI 與資料中心的成長敘事，都會遇到現實瓶頸。
在這個脈絡下，SMR 看起來確實像是一個值得關注的解法。它有機會提供較穩定的基載電力， 並且在資料中心與工業用電場景裡，成為風電、太陽能與儲能之外的另一條長期選項。 對算力基礎設施而言，能源來源的可靠性，很多時候比理論上的最便宜成本更重要。
但我認為這件事現在還不能只看「SMR 很有未來」這一句話。SMR 是否能真正成為解法， 很大一部分仍取決於燃料供應鏈，特別是 HALEU。過去商業可得的 HALEU 供應高度依賴俄羅斯， 這使得技術前景與地緣政治風險被綁在一起，也讓整個產業需要更長時間觀察。
English
If we accept one premise, that AI is built on compute, then the next question becomes very direct: what is compute built on? The answer is not the model itself, but energy. Without sufficient, stable, and expandable power, every growth narrative around large models, inference services, edge AI, and data centers will eventually run into a physical constraint.
In that context, SMRs do look like a solution worth watching. They may provide steadier baseload power and become another long-term option, alongside wind, solar, and storage, for data centers and industrial power users. For compute infrastructure, the reliability of the energy source is often more important than the theoretically cheapest cost on paper.
But I do not think this can be reduced to the simple statement that SMRs are the future. Whether SMRs can truly become a solution depends in large part on the fuel supply chain, especially HALEU. Commercial HALEU availability has long been closely tied to Russia, which means technological promise and geopolitical risk are still linked together. That alone makes this a long-duration observation theme rather than an easy conclusion.
COMMENTARY POINTS
- AI growth ultimately runs into an energy question
- SMRs may help supply stable power for future compute demand
- HALEU remains a strategic bottleneck in the fuel chain
- Centrus Energy is a useful long-term U.S. supply-chain marker
Chinese
這也是為什麼美國本土的 HALEU 供應進展值得看。`Centrus Energy`（美股代碼 `LEU`） 目前的重要性，不只是因為它是核燃料公司，而是因為它在美國本土 HALEU 供應鏈裡， 已經具備先行者位置。根據 DOE 與公司公開資訊，Centrus 已在俄亥俄州建立示範產線， 並向美國能源部交付產品，這讓它成為很多下一代反應爐與 SMR 故事中繞不開的一環。
但這裡我更傾向把 `LEU` 當成長期觀察名單，而不是短線結論。原因很簡單： 一家公司在供應鏈上的戰略位置很重要，但產業何時真正放量、政策支持能否持續、 本土擴產速度是否跟得上，以及 SMR 商業化時程是否如預期推進，這些都還有很多變數。
所以我目前的看法比較簡單：SMR 可能是算力能源問題的解法之一，但燃料供應仍是核心瓶頸； `Centrus Energy` 的進展值得持續追蹤，但這是一個需要時間驗證的長期主題， 而不是今天就能下結論的單線敘事。
English
That is exactly why U.S. domestic progress on HALEU matters. `Centrus Energy` (`LEU`) is important not simply because it is a nuclear fuel company, but because it occupies an early strategic position in the U.S. HALEU supply chain. Based on public DOE and company disclosures, Centrus has established a demonstration production line in Ohio and delivered product to the U.S. Department of Energy, making it difficult to ignore in many next-generation reactor and SMR discussions.
Even so, I would rather treat `LEU` as a long-term watchlist name than a short-term conclusion. Strategic positioning in a supply chain matters, but there are still many moving parts: when the industry truly scales, whether policy support remains durable, how fast domestic production can expand, and whether SMR commercialization timelines progress as expected.
So my current view is straightforward: SMRs may be one of the answers to AI's energy problem, but fuel supply is still the central bottleneck. `Centrus Energy` is worth following closely, but this is a theme that needs to be tested over time rather than accepted as a finished story today.