来源 | 新智元 「光刻机巨头」入局AI了！ 路透独家，ASML豪掷15亿美元领投Mistral AI，将成为这家法国AI新星的最大股东。 本次C轮融资总额约20亿美元，融资完成后，Mistral AI将成为欧洲最有价值的AI公司。 这波操作，堪称欧洲科技圈的「王炸」。 这笔天价融资，将为Mistral AI的研发注入一针强心剂。 光 刻 机 巨 头 ， 看 上 了 A I 提起ASML，无人不晓。 它可是全球芯片制造的「幕后推手」，独家供应EUV（极紫外光刻机），一台售价1.8亿美元。 台积电、英特尔等顶尖芯片代工厂，全都得靠EUV生产最尖端的芯片。 这次ASML不仅砸下重金，还能拿下Mistral AI的董事会席位，堪称强强联手。 业内人士透露，这笔投资不仅是财务支持，更是欧洲科技主权的一次战略布局。 那么，ASML为何如此大手笔？答案就藏在了它的「生意经」里。 作为全球芯片制造的命脉，EUV光刻机是生产尖端芯片的「神器」，但也在不断用AI优化设备性能。 Mistral AI的数据分析和AI能力，正好能帮ASML在芯片制造的精度和效率上更进一步，从而开发更多的产品。 如今，这家唯一能与OpenAI相匹敌的 ...

Core Viewpoint - The rapid development of AI technology is leading to significant environmental costs, including increased carbon emissions and resource consumption, prompting companies like Microsoft to invest heavily in carbon offset initiatives [3][5][10]. Group 1: Microsoft's Investment in Carbon Offsetting - Microsoft has signed a deal with Vaulted Deep to achieve a carbon removal target of 4.9 million tons by 2038, with the total value of the deal reportedly exceeding $1 billion, potentially costing Microsoft up to $1.7 billion [6][7]. - Vaulted Deep specializes in converting organic waste into a "bio-slurry" that is injected deep underground to permanently remove carbon and reduce methane emissions [8]. - This expenditure is essentially a purchase of "emission rights" for Microsoft, as the company faces increasing carbon emissions due to its expanding cloud computing and AI operations [9][10]. Group 2: Environmental Impact of AI - The environmental costs of AI are often overlooked, with current carbon accounting mechanisms being criticized for their inadequacies [11][12]. - Major tech companies, including Amazon, Microsoft, Alphabet, and Meta, have seen their indirect carbon emissions increase by an average of 150% from 2020 to 2023 due to the high energy demands of data centers [13]. - Specific data shows that Microsoft's operational carbon emissions increased by 155% in 2023 compared to three years prior [14]. Group 3: Water and Resource Consumption - AI models, such as Mistral Large 2, have significant water consumption, using 281,000 cubic meters of water over 18 months, equivalent to filling 112 Olympic swimming pools [16]. - The lifecycle of AI models shows that training and inference phases account for 85.5% of greenhouse gas emissions and 91% of water consumption [17]. Group 4: Future Projections and Industry Response - By 2027, AI is projected to consume up to 6.6 billion cubic meters of water, double Switzerland's annual consumption, and data center energy consumption may double by 2030 [20]. - The tech industry is under pressure to reduce emissions, with companies like Google, Microsoft, and Meta committing to net-zero emissions by 2030, while Amazon aims for 2040 [23]. - Experts emphasize the need for companies to prioritize renewable energy and transparency in environmental impact disclosures, with government intervention being crucial for effective regulation [27][28].

Core Insights - The article discusses the environmental costs associated with the rapid development of artificial intelligence (AI), highlighting Microsoft's significant investment in carbon removal initiatives as a response to its increasing carbon footprint [3][4][6]. Group 1: Microsoft's Investment in Carbon Removal - Microsoft has signed a deal with Vaulted Deep to achieve a carbon removal target of 4.9 million tons by 2038, with the total value of the contract reportedly exceeding $1 billion, potentially costing Microsoft up to $1.7 billion [4][6]. - The method employed by Vaulted Deep involves converting organic waste into a "bio-sludge" and injecting it deep underground to permanently remove carbon and prevent pollution [5][6]. Group 2: AI's Environmental Impact - The carbon emissions from AI operations are substantial, with a report indicating that the training and operation of Mistral AI's flagship model generated CO2 emissions equivalent to that of 5,000 cars over 18 months [3][8]. - Major tech companies, including Microsoft, have seen significant increases in their carbon emissions, with Microsoft's emissions rising by 23.4% since 2020 due to the expansion of its cloud and AI businesses [6][8]. Group 3: Criticism of Carbon Offset Strategies - Experts criticize Microsoft's approach of purchasing carbon credits as a means to offset emissions, arguing that it does not address the root cause of emissions and can mislead the public regarding actual reductions [6][12]. - The current carbon accounting mechanisms are seen as flawed, allowing companies to claim zero emissions through market-based accounting without making substantial changes to their energy consumption practices [7][12]. Group 4: Future Projections and Industry Actions - By 2027, AI is projected to consume up to 6.6 billion cubic meters of water annually, and data center energy consumption could double by 2030, surpassing the total energy consumption of countries like Japan [12][13]. - Major tech companies are taking steps towards sustainability, with commitments to achieve net-zero emissions by 2030 or 2040, but there is a call for more transparency and government regulation to ensure accountability [13][14].

Core Insights - The article discusses the environmental costs associated with the rapid development of artificial intelligence (AI), highlighting Microsoft's significant investment in carbon removal initiatives as a response to its increasing carbon footprint [2][7]. Group 1: Microsoft's Investment in Carbon Removal - Microsoft has signed a deal with Vaulted Deep to achieve a carbon removal target of 4.9 million tons by 2038, with the total value of the contract reportedly exceeding $1 billion, potentially costing Microsoft up to $1.7 billion based on a fee of $350 per ton [3][5]. - The investment is part of Microsoft's strategy to purchase "emission rights" as its carbon emissions have increased by 23.4% since 2020 due to the expansion of its cloud computing and AI businesses [7]. Group 2: Environmental Impact of AI - The AI industry is facing growing scrutiny over its environmental impact, with reports indicating that the indirect carbon emissions from major tech companies like Amazon, Microsoft, Alphabet, and Meta have increased by an average of 150% from 2020 to 2023 [9][10]. - A specific AI model, Mistral Large 2, produced 20.4 tons of CO2 over 18 months, equivalent to the annual emissions of 5,000 average cars, and consumed 281,000 cubic meters of water [10]. Group 3: Criticism of Carbon Offset Strategies - Experts criticize Microsoft's approach of "buying offsets," arguing that it distracts from the need for genuine emission reductions at the source [8][14]. - The current carbon accounting mechanisms are seen as flawed, allowing companies to claim zero emissions through market-based accounting, which may not reflect actual environmental impacts [9][10]. Group 4: Future Projections and Industry Actions - By 2027, AI is projected to consume up to 6.6 billion cubic meters of water annually, and data center energy consumption could double by 2030, surpassing the total energy consumption of countries like Japan [13][14]. - Major tech companies are making commitments to achieve net-zero emissions by 2030 or 2040, but experts emphasize the need for a focus on renewable energy and reducing emissions at the source rather than relying solely on carbon credits [14][15].

在华盛顿州雷德蒙德市的微软总部，工程师们的手指正在键盘上翻飞，为人工智能（AI）的未来编写着一行行代码。 与此同时，在1500公里外，一场看似毫不相干的"怪事"正在发生：在堪萨斯州地下深处，巨型泵机正将混合着人类粪便和污水的黏稠"生物泥浆"，源源不断 地注入地壳深处。 这听起来匪夷所思的"埋粪"操作，正是微软为AI飞速发展所付出的"环境账单"。 AI的"超能力"光环背后，是惊人的电力、水资源消耗和飙升的碳排放。法国AI初创公司Mistral AI的最新报告显示，其旗舰模型Mistral Large 2训练和运行18 个月，产生的二氧化碳竟然相当于5000辆家用轿车一年的排放量。 每经记者｜兰素英 每经编辑｜高涵 在AI基础设施规模扩张时，科技公司对自身碳足迹的披露却依然不甚透明。"埋粪"不过是冰山一角，"绿色焦虑"正逐步成为AI产业链上无法回避的巨大成 本。 图片来源：视觉中国-VCG41N2155090857 最近，微软和一家名为Vaulted Deep的生物科技初创公司签了一份大单。根据协议，Vaulted Deep将代表微软在2038年前完成490万吨的碳去除目标。 虽然具体金额没公布，但有报道称这笔交 ...

图片来源： Mistral AI Mistral AI ，这家法国公司是 AI 助手 Le Chat 和多个基础模型背后的力量组成， 被官方视为法国最有前途的科技初创企业之一，且可以说是能与 OpenAI 竞争的欧洲公司。 然而，与其 60 亿美元的估值相比，其全球市场份额仍相对较低。 然而，其聊天助手近期在移动应用商店的发布引发了一些热议，尤其是在其本土国家。 法国总统埃马纽埃尔 ·马克龙在巴黎人工智能行动峰会前的一次电视采访中表示："去下载由 Mistral 开发的 Le Chat ，而不是 OpenAI 的 ChatGPT ——或者 其他什么。" 尽管这波关注可能令人鼓舞， Mistral AI 在与 OpenAI 等公司竞争时仍面临挑战，同时还要坚持其作为"全球最环保、领先的独立 AI 实验室"的自我定位。 什么是 Mistral AI ？ 自 2023 年成立以来， Mistral AI 已筹集了大量资金，其目标是"让前沿 AI 触手可及"。虽然这不是直接针对 OpenAI 的批评，但这一口号旨在强调公司对 AI 开放性的倡导。 图片来源： Mistral AI 其替代 ChatGPT 的聊天助手 ...