ITU 報告聚焦數位業用電與排放集中，亞洲科技供應鏈須核對電力來源與營運邊界

國際電信聯盟與世界基準聯盟檢視 200 家數位企業的 2024 年資料，Eco-Business 報導其中前 50 大營運排放者占樣本 92%，前 10 大約排放 1.63 億噸二氧化碳當量，占 54%。在 163 家揭露用電的企業中，總用電約 494 TWh，前 10 大使用者占 269 TWh。報告凸顯少數企業具有高度影響力，但排名不等於個別設施或產品的完整碳足跡。

The International Telecommunication Union and World Benchmarking Alliance reviewed 2024 data for 200 digital companies. Eco-Business reports that the 50 largest operational emitters represented 92% of the sample, while the top 10 produced about 163 million tonnes of carbon dioxide equivalent, or 54%. Among 163 companies disclosing electricity, total use was about 494 TWh and the top 10 used 269 TWh. Concentration shows influence, but a ranking is not a complete footprint for each facility or product.

數位業的能源邊界跨越資料中心、電信網路、半導體與電子製造、物流、辦公室及冷媒。Eco-Business 指出東亞與太平洋企業的再生電力占比約 18.5%，北美約 68%，兩區合計占樣本逾 80% 用電。Reuters 先前報導主要科技企業因人工智慧擴張而使間接排放快速上升，並報導歐盟要求一定規模資料中心揭露能源與水效率。

Digital-sector energy boundaries span data centres, telecom networks, semiconductor and electronics plants, logistics, offices and refrigerants. Eco-Business reports renewable electricity shares of about 18.5% for East Asia and the Pacific and 68% for North America, with the two regions representing more than 80% of disclosed electricity. Reuters previously documented rising indirect emissions among major technology firms as artificial intelligence expanded and later reported an EU requirement for larger data centres to disclose energy and water efficiency.

年度再生能源憑證或購電合約不必然證明資料中心在實際運轉的地點與時段使用低碳電力。企業還可能以不同組織邊界、位置基礎與市場基礎方法揭露 Scope 2，使跨公司比較失真。人工智慧工作負載、備援發電、冷卻、用水與晶片製造又分散在多個資產和供應商，若沒有電表、時間戳、地點、契約與分配方法，採購宣稱難以連回實際服務。

Annual renewable certificates or power-purchase agreements do not necessarily prove that a data centre used low-carbon electricity at the place and hour of operation. Companies may also use different organisational boundaries and location-based or market-based Scope 2 methods. AI workloads, backup generation, cooling, water and chip production are distributed across assets and suppliers. Without meters, timestamps, locations, contracts and allocation methods, procurement claims cannot be connected reliably to the delivered service.

資料中心營運商應以設施為單位保存主電表、分項計量、負載曲線、電網排放因子、再生電力合約、憑證序號與註銷紀錄，並分列位置基礎與市場基礎結果。雲端與 AI 服務採購方應要求供應商提供服務區域、運算時段、工作負載分配、PUE、備援燃料與分攤方法。晶片及設備供應商則須把晶圓廠電力、製程氣體、產量與產品碳足跡版本連結到交付批次。

Data-centre operators should retain facility-level main meters, submetering, load curves, grid factors, renewable contracts, certificate serials and retirement records, while reporting location-based and market-based results separately. Cloud and AI buyers should require service region, computing hours, workload allocation, PUE, backup fuel and allocation methods. Chip and equipment suppliers should connect fab electricity, process gases, output and product-carbon-footprint versions to delivered production lots.

三筆固定來源互相補足。Eco-Business 提供 ITU 與 WBA 報告的樣本、用電、排放集中度及區域再生電力差異。Reuters 的 2025 年報導記錄人工智慧擴張下主要科技企業間接排放上升；Reuters 的 2026 年 9 月報導則說明歐盟將要求 500 kW 以上資料中心揭露能源與用水效率。證據支持揭露壓力正在由年度公司數字下沉到設施效率，但不能取代未公開的原始電表與合約資料。

Three fixed-scope reports provide complementary evidence. Eco-Business covers the ITU and WBA sample, energy and emissions concentration, and regional renewable-electricity differences. Reuters reported rising indirect emissions at major technology companies during AI expansion in 2025. A September 2026 Reuters report explains the EU's planned energy and water efficiency disclosure for data centres above 500 kW. The evidence shows reporting moving toward facility efficiency, but it cannot replace underlying meter and contract records that remain non-public.

科技業的電力證據將直接影響客戶 Scope 3、綠色採購、資料中心選址與融資。亞洲電網再生能源供給與市場工具的差異，使相同年度憑證量可能對應不同實體電力條件。電信商、雲端平台、晶圓廠和伺服器供應商需要共同定義組織與服務邊界，否則一筆能源屬性可能在多個產品聲明中被重複使用。

Electricity evidence in technology supply chains affects customer Scope 3, green procurement, data-centre location and finance. Differences in Asian grid supply and procurement tools mean the same annual certificate volume may correspond to different physical conditions. Telecom operators, cloud platforms, fabs and server suppliers need common organisational and service boundaries, or one energy attribute could be reused across multiple product claims.

SNN editorial analysis：全球制度變化是 ITU 與 WBA 將數位業用電與營運排放的集中度，以及電力的時段與地點匹配，擴大為可比較的市場要求；這項制度要求會透過雲端採購、AI 服務合約、半導體客戶碳資料與資料中心融資傳導到台灣資料中心營運商、電信商及晶片供應鏈，使台灣市場必須證明服務用電與產品碳資料的共同邊界。第一，台灣資料中心能源與設施主管應在每月結算及年度揭露前，比對主電表、分項負載曲線、PUE、備援燃料紀錄、電力合約、憑證序號與註銷時間，並在設施或租戶邊界改變時重算位置基礎與市場基礎 Scope 2。第二，台灣雲端與 AI 採購及法務主管應在簽約、容量擴充與續約時，核對服務區域、運算時段、工作負載分配、碳資料方法、供應商查證聲明與服務水準紀錄，未取得可追溯資料前不得宣稱工作負載已達 24/7 無碳。第三，台灣半導體製造與產品碳管理主管應在客戶出貨及季度產品碳足跡更新前，比對晶圓批次、廠區電表、製程氣體、產量分母、能源屬性分配與第三方確信紀錄，並在製程或電力組合變更時發行新版本。這是揭露前證據基礎設施的條件式台灣傳導分析，不代表報告中的排名可直接套用到單一台灣產品，也不等於年度再生能源採購證明每小時實體用電。

SNN editorial analysis: The global institutional change is the ITU and WBA expansion of concentrated digital-sector energy and operational emissions, including time-and-location matching of electricity, into a comparable market requirement. This requirement can flow through cloud procurement, AI service contracts, semiconductor customer carbon data and data-centre finance to Taiwan data-centre operators, telecom companies and chip supply chains, requiring the Taiwan market to evidence a common boundary for service electricity and product carbon data. First, before monthly settlement and annual disclosure, Taiwan data-centre operators and facilities managers should reconcile main meters, submeter load curves, PUE, backup-fuel logs, power contracts, certificate serials and retirement timing, and must recalculate location-based and market-based Scope 2 when facility or tenant boundaries change. Second, before any contract, capacity expansion or renewal, Taiwan cloud and AI procurement leaders and legal leaders should verify service-region records, compute-hour logs, workload-allocation data, the carbon-data method, supplier assurance statement and service-level record, and must not claim 24/7 carbon-free operation without traceable evidence. Third, before customer shipment and each quarterly product-footprint update, Taiwan semiconductor manufacturers and product data owners should compare wafer-lot records, fab meters, process-gas data, production denominators, energy-attribute allocation and third-party assurance statements, and must issue a new version when the process or power mix changes. This Pre-Disclosure Evidence Infrastructure interpretation does not apply the report ranking directly to one Taiwan product, and annual renewable procurement is not proof of physical electricity in every hour.

後續應追蹤企業是否揭露設施層級用電、水耗、PUE、位置基礎與市場基礎排放，以及 24/7 無碳電力的時段匹配。歐盟資料中心標示制度、AI 容量成長、亞洲電力市場與儲能投資，將決定數位企業能否把年度承諾轉成可核對的營運證據。

Future review should track facility-level electricity, water, PUE, location-based and market-based emissions, and hourly matching for 24/7 carbon-free energy. EU data-centre labels, AI capacity growth, Asian power-market reforms and storage investment will determine whether annual commitments become verifiable operating evidence.