AI News AI资讯 2h ago Updated 1h ago 更新于 1小时前 49

Datacentres drive up big tech’s carbon emissions to a third of those of France 数据中心推动大型科技公司碳排放量升至法国的三分之一

Microsoft, Amazon, and Google collectively emitted 119 million metric tonnes of CO2 equivalent in the fiscal year ending March 2026, representing a nearly 20% increase driven by AI infrastructure expansion. This combined carbon footprint is approximately one-third of France's total annual emissions, highlighting the massive environmental scale of the current AI boom. Emissions rose due to datacenter construction and supply chain activities, with Microsoft seeing a 25% increase, Google an 18% inc 微软、亚马逊和谷歌去年的碳排放总量达1.19亿吨二氧化碳当量,同比增长近20%,相当于法国总排放量的三分之一。 碳排放激增的主要驱动力是AI热潮引发的数据中心建设热潮及相关的供应链活动扩张。 专家批评云服务商的“绿色营销”策略,指出其通过外包碳足迹掩盖了实际环境影响,且全球碳信用市场可能面临供应短缺。 尽管排放上升,三大巨头仍坚持净零目标(微软和谷歌2030年,亚马逊2040年),但面临物理基础设施与虚拟碳信用双重资源瓶颈。

75
Hot 热度
70
Quality 质量
65
Impact 影响力

Analysis 深度分析

TL;DR

  • Microsoft, Amazon, and Google collectively emitted 119 million metric tonnes of CO2 equivalent in the fiscal year ending March 2026, representing a nearly 20% increase driven by AI infrastructure expansion.
  • This combined carbon footprint is approximately one-third of France's total annual emissions, highlighting the massive environmental scale of the current AI boom.
  • Emissions rose due to datacenter construction and supply chain activities, with Microsoft seeing a 25% increase, Google an 18% increase, and Amazon a 16% increase.
  • Experts warn that cloud providers are effectively outsourcing and obscuring the carbon footprints of other corporations migrating to their platforms.
  • A potential shortage of high-quality carbon credits is emerging as a bottleneck for tech giants attempting to offset their growing emissions while maintaining net-zero pledges.

Why It Matters

This trend signals a critical conflict between the rapid scaling of AI infrastructure and corporate sustainability goals, challenging the narrative that cloud computing is inherently green. For industry stakeholders, it underscores the urgent need for transparency in Scope 3 emissions and the limitations of relying on carbon offsets to mitigate the environmental impact of massive datacenter builds.

Technical Details

  • Emission Metrics: The three tech giants emitted 119m mTCO₂e, up from 101m mTCO₂e the previous year. Microsoft alone accounted for 20m mTCO₂e (a 25% rise), while Amazon and Google reported 16% and 18% increases respectively.
  • Infrastructure Costs: Global tech companies are projected to spend $765 billion in the current year, primarily directed toward constructing AI datacenters in locations ranging from Norway to North Tyneside.
  • Energy Consumption: New datacenter projects announced recently are estimated to consume 1.3% of global electricity usage, nearly doubling current datacenter demand, with the majority of this load expected in the US.
  • Market Constraints: Reports indicate a tightening supply of carbon credits, suggesting that the virtual market for offsets may struggle to meet the volume required by major tech firms to achieve net-zero targets.
  • Future Projections: JLL estimates approximately 1,200 new datacenters will be built globally by 2030, with demand overwhelmingly driven by AI workloads.

Industry Insight

  • Re-evaluate Net-Zero Strategies: Companies must move beyond simple carbon offsetting and invest heavily in direct decarbonization technologies, such as renewable energy integration and efficient hardware, as offset markets become scarce and scrutinized.
  • Supply Chain Transparency: Organizations relying on cloud services should recognize that their own carbon footprints are being shifted to hyperscalers; therefore, selecting cloud partners based on verified, granular sustainability metrics is becoming a compliance and reputational necessity.
  • Infrastructure Planning: The severe strain on global power grids and carbon markets indicates that future AI development will face hard limits on energy availability and cost, necessitating more energy-efficient model architectures and localized data processing strategies.

TL;DR

  • 微软、亚马逊和谷歌去年的碳排放总量达1.19亿吨二氧化碳当量,同比增长近20%,相当于法国总排放量的三分之一。
  • 碳排放激增的主要驱动力是AI热潮引发的数据中心建设热潮及相关的供应链活动扩张。
  • 专家批评云服务商的“绿色营销”策略,指出其通过外包碳足迹掩盖了实际环境影响,且全球碳信用市场可能面临供应短缺。
  • 尽管排放上升,三大巨头仍坚持净零目标(微软和谷歌2030年,亚马逊2040年),但面临物理基础设施与虚拟碳信用双重资源瓶颈。

为什么值得看

这篇文章揭示了AI产业爆发式增长背后的环境代价,打破了科技巨头“绿色云计算”的叙事神话,为评估AI可持续发展的真实性提供了关键数据支撑。对于政策制定者和投资者而言,它警示了算力基础设施扩张对全球能源结构和碳市场的巨大冲击,强调了监管和技术伦理的重要性。

技术解析

  • 排放数据对比:截至2026年3月的财年,三大科技公司集体排放1.19亿吨CO₂e,较前一年的1.01亿吨显著增加。其中微软排放2000万吨(+25%),谷歌18%增幅,亚马逊16%增幅(供应链排放增20%)。
  • 驱动因素分析:排放增长主要归因于AI训练和推理所需的数据中心基础设施扩建。JLL预测到2030年全球将新建约1200个数据中心,Uptime Institute估计去年宣布的大型项目将消耗全球1.3%的电力,使数据中心能耗接近翻倍。
  • 碳抵消机制局限:微软报告暗示全球碳信用市场供应不足,难以满足科技巨头快速扩张的抵消需求。这表明仅靠购买碳信用无法完全解决由物理基础设施扩张带来的真实排放增长。
  • 长期目标现状:尽管短期排放反弹,公司仍维持长期净零承诺。谷歌声称其AI解决方案已帮助外部减少4100万吨排放,试图以“AI减碳效益”对冲自身“AI增碳成本”。

行业启示

  • 重新定义AI可持续性:行业需从单纯的“使用绿色能源”转向全生命周期碳管理,正视数据中心建设(Scope 3排放)带来的巨大碳足迹,避免陷入“漂绿”争议。
  • 基础设施与能源约束:AI发展的瓶颈不仅是算法和数据,更是电力供应和土地/水资源。企业需提前布局能源基础设施,并探索更高效的冷却技术和硬件设计以降低单位算力能耗。
  • 监管与透明度升级:随着碳信用市场供需失衡,政府和企业应建立更严格的披露标准,区分范围1、2、3排放,并推动实体减排而非依赖虚拟信用,以确保气候目标的真实性。

Disclaimer: The above content is generated by AI and is for reference only. 免责声明:以上内容由 AI 生成,仅供参考。

GPU GPU Training 训练 Policy 政策