書面答覆 · 2024-01-09 · 屆國會 14
AI算力與能源可持續平衡
議員質詢政府如何在推動國家AI戰略2.0所需的算力增長與能源可持續目標之間取得平衡。政府回應強調通過提升資料中心能效、推廣液冷技術、制定綠色標準及支援綠色計算方法，確保AI基礎設施環保可持續。核心爭議在於如何兼顧AI發展需求與氣候承諾的矛盾。
關鍵要點
- • 提升資料中心能效
- • 推廣液冷節能技術
- • 發展綠色計算方法
推動綠色高效AI算力發展
關注算力與能源平衡
加強綠色資料中心建設
“The Government is committed to anchoring the necessary compute power and growing the DC sector in a sustainable manner.”
參與人員 (2)
完整譯文（中文）
Hansard 原始記錄 · 2026-05-02
47 林偉傑醫生向通訊及資訊部長提問，關於國家人工智慧（AI）戰略2.0，(a) 政府如何計劃在滿足AI發展所需的計算能力增長與新加坡的能源可持續發展目標之間取得平衡；(b) 有哪些舉措或策略確保支援AI增長所需的基礎設施和計算能力保持環境可持續性。
張玉娟女士：我們國家人工智慧戰略（NAIS）2.0的重要推動力之一，是設於新加坡的資料中心（DC）內的人工智慧計算能力。政府致力於以可持續的方式錨定必要的計算能力並發展資料中心行業，這與我們的國際氣候承諾保持一致。
一項關鍵策略是提升資料中心的能源效率並促進高效冷卻解決方案的部署。例如，AI計算基礎設施通常依賴液冷技術，相較於空氣冷卻，液冷在處理高強度工作負載時更節能。其他措施包括：(a) 通過經濟發展局的排放資源效率補助金支援資料中心運營商減少溫室氣體排放；(b) 制定可持續發展標準，如資訊通訊媒體發展局（IMDA）的熱帶資料中心標準，使資料中心能夠在更高溫度下安全執行並減少冷卻能耗；(c) 審查建築與建設局（BCA）與IMDA聯合推出的資料中心綠色標誌認證計劃，更新能源效率標準。
除了可持續的AI計算基礎設施外，投資綠色計算方法的發展也非常重要。這些方法包括：(a) 編碼和演算法最佳化，(b) 軟體與硬體的最佳化，(c) 制定低資料量、低能耗AI模型的標準。政府將繼續深化與研究界和產業夥伴的國際及國內合作。
英文原文
SPRS Hansard · Fetched: 2026-05-02
47 Dr Lim Wee Kiak asked the Minister for Communications and Information with regard to the National Artificial Intelligence (AI) Strategy 2.0 (a) how does the Government plan to balance the increased computing power necessary for AI developments with Singapore's energy sustainability goals; and (b) what initiatives or strategies are in place to ensure that the infrastructure and computing power needed for AI growth remain environmentally sustainable.
Mrs Josephine Teo : An important enabler of our National Artificial Intelligence Strategy (NAIS) 2.0 strategy is artificial intelligence (AI) compute power housed in data centres (DCs) in Singapore. The Government is committed to anchoring the necessary compute power and growing the DC sector in a sustainable manner, consistent with our international climate commitments.
One key strategy is to improve the energy efficiency of DCs and facilitate the deployment of efficient cooling solutions. For example, AI compute infrastructure typically relies on liquid cooling, which is more energy-efficient than air cooling for such intensive workloads. Other measures include: (a) supporting DC operators to reduce greenhouse gas emissions through the Economic Development Board's Resource Efficiency Grant for Emissions; (b) developing sustainability standards, such as the Infocomm Media Development Authority's (IMDA)'s Tropical DC Standards that enable DCs to operate safely at higher temperatures and use less energy for cooling; and (c) reviewing Building and Construction Authority (BCA)-IMDA's Green Mark for DCs certification scheme to update the energy-efficiency criteria.
Beyond sustainable AI compute infrastructure, it is important to invest in the development of green computing methods. These methods include (a) coding and algorithm optimisation, (b) software-hardware optimisation, and (c) developing standards for low-data, low-energy AI models. The Government will continue to deepen international and domestic partnerships with the research community and industry partners on this front.