How Data Centers Broke American Politics

数据中心是如何撕裂美国政治的

Joe Allen wants to know what the robot’s name is. At first, I think the double whiskey we downed at the hotel bar must be getting to him, as he sprints like a kid at recess toward a delivery vehicle. I’m anxious about the time: I’m supposed to be driving Allen to the airport, my Hail Mary effort to get him alone for a proper conversation before he leaves Atlanta.

乔·艾伦（Joe Allen）想知道那个机器人的名字。起初，我以为我们在酒店酒吧喝下的双份威士忌让他有些上头，因为他像个课间休息时的孩子一样，朝着一辆送货车飞奔而去。我为时间感到焦虑：我本该开车送艾伦去机场，这是我在他离开亚特兰大前，能与他单独进行深入交谈的最后一次“孤注一掷”的努力。

I’ve been back and forth with Allen, who rose to national prominence as the anti-tech prophet on Steve Bannon’s War Room podcast, for a few weeks now, seeking an entry into the growing movement against AI. I’d come down to Georgia, a hot spot of data center resistance, to hear one of Allen’s lectures, which he’s been giving at locations across the country. But when I got to the event, he told me he had to leave immediately after.

几周以来，我一直与艾伦保持联系。他因在史蒂夫·班农（Steve Bannon）的《战情室》（War Room）播客中担任反科技先知而声名鹊起，我一直在寻求切入点，以了解日益壮大的反人工智能运动。我来到佐治亚州——这个反数据中心运动的热点地区——是为了听艾伦在全国各地巡回举办的讲座之一。但当我赶到活动现场时，他却告诉我，活动一结束他就必须立刻离开。

Fortunately, his flight keeps getting pushed back, which means I can follow him around on various side quests—like this one, to meet some friends at a nearby restaurant. He manages to evade most of my questions about their identities, changing the topic to more interesting things, like the “inverse singularity” (his idea that humanity will approach a point where our intellect and creativity collapse as we rely increasingly on AI for information). Allen’s friends at the restaurant have similarly little information about me, it seems. On our way, one of them calls Allen to ask for an ETA. “I’m bringing a friend—well, a frenemy,” he says, giving me a dramatic side-eye.

幸运的是，他的航班一再延误，这意味着我可以跟随他去处理各种“支线任务”——比如现在，去附近的餐厅见几个朋友。他设法回避了我关于他们身份的大多数问题，转而谈论更有趣的话题，比如“逆奇点”（他的观点是：随着人类越来越依赖人工智能获取信息，人类的智力和创造力将走向崩溃）。餐厅里的艾伦的朋友们似乎对我知之甚少。在路上，其中一人打电话给艾伦询问预计到达时间。“我带了个朋友——嗯，准确说是‘敌友’，”他一边说，一边夸张地斜了我一眼。

There are delivery robots all over downtown Atlanta, one of a handful of major metropolitan areas where they’re being rolled out to bring people burgers or tacos. Allen catches up to this one—which, I’m alarmed to see, has luminous white lights shaped like eyes on its front—and reads a name off the side: “CJ.” I think of the title of Allen’s recent book, Dark Aeon: Transhumanism and the War Against Humanity, and wonder what happened to the delivery worker that this robot replaced. Its eyes look pathetically, limply, past me.

亚特兰大市中心到处都是送货机器人，这里是少数几个开始推广机器人送汉堡或塔可（tacos）的大都市之一。艾伦追上了这台机器人——令我惊恐的是，它的正面有发光的白色眼睛状灯光——并读出了侧面的名字：“CJ”。我联想到艾伦最近那本书的书名——《黑暗永世：超人类主义与反人类战争》（Dark Aeon: Transhumanism and the War Against Humanity），不禁好奇被这台机器人取代的送货员现在怎么样了。它的眼睛显得可怜而无力，越过我看向远方。

When people ask what I’m doing in Georgia, I give them something easy. People hate data centers, I say, so I’m trying to figure out if they hate AI too. That’s all true. But the topic has also consumed me. In all my years of reporting on energy and the environment, I have never seen anything like the type of vitriol leveled at data centers, or the near-religious fury that has built up around their construction, or the way this resistance seems to be uniting people from the complete opposite ends of the political spectrum.

当人们问我在佐治亚州做什么时，我给出了一个简单的回答。我说，人们讨厌数据中心，所以我想弄清楚他们是否也讨厌人工智能。这些都是事实。但这个话题也让我深陷其中。在我报道能源和环境问题的这些年里，我从未见过像针对数据中心那样的尖刻抨击，也从未见过围绕其建设所产生的近乎宗教般的愤怒，更没见过这种抵制运动竟能将政治光谱两端的人们团结在一起。

As I’m writing this piece, a friend sends me a post from a “white Christian nationalism” account on X: a seemingly AI-generated cartoon of Smokey Bear standing in front of a burning data-center-like structure, with the text “ONLY YOU CAN PREVENT DATA CENTERS.” What is happening? my friend asks. I’ll try to explain.

在我写这篇文章时，一位朋友在 X 上发给我一个“白人基督教民族主义”账号的帖子：那是一幅看似由人工智能生成的卡通画，画中“斯莫基熊”（Smokey Bear）站在一座燃烧的、类似数据中心的建筑前，配文写着：“只有你能阻止数据中心。”我朋友问，这到底是怎么了？我试着解释一下。

Once upon a time, you may have been living next to a data center and never heard anything about it. For the past few decades, data centers have been quietly built all over the US, largely in areas with both fiber connectivity and proximity to power—Silicon Valley and the DC area were popular choices. They helped stream Netflix or store stuff in the cloud or do any one of the countless stupid and important things we all do constantly online.

曾几何时，你可能就住在数据中心隔壁，却对此一无所知。在过去的几十年里，数据中心在全美各地悄然兴起，主要选址在既有光纤连接又靠近电力供应的地区——硅谷和华盛顿特区周边是热门选择。它们帮助我们播放 Netflix、存储云端数据，或者处理我们在网上不断进行的无数琐碎而重要的事情。

Then came ChatGPT and the AI race. Suddenly, tech oligarchs argued, we needed to build as many data centers as possible—and fast. Making the best AI, they said in speeches and news interviews and social media posts, required a massive buildout, especially of hyperscalers—behemoth versions of the structures we already had. The Trump administration bought in, rolling back environmental regulations to get the government out of the way of development.

随后，ChatGPT 和人工智能竞赛开始了。科技寡头们突然宣称，我们需要尽可能多地建设数据中心，而且要快。他们在演讲、新闻采访和社交媒体帖子中表示，要打造最强的人工智能，就需要大规模的扩建，尤其是“超大规模数据中心”（hyperscalers）——即我们现有设施的巨型版本。特朗普政府对此深信不疑，并放宽了环境法规，以扫清政府在发展道路上的障碍。

Data centers began cropping up everywhere, as companies hurried to find the cheapest land, power, and water. Every day brought a splashier headline. Elon Musk reportedly trucked in unpermitted gas turbines to his Memphis xAI campus, located in a majority-Black neighborhood already suffering from high levels of pollution and contamination. A popular (later disputed) Washington Post article estimated that a single ChatGPT search guzzled the equivalent of a bottle of water. People eyed their electric bills as new data center loads caused utilities to forecast sky-high demand. If you trusted various economic outlets, data center construction now underpinned pretty much the entire US economy. All of this as tech CEOs publicly mused about how AI would take everyone’s jobs and replace human creativity itself.

随着公司争相寻找最廉价的土地、电力和水资源，数据中心开始在各地涌现。每天都有更引人注目的头条新闻。据报道，埃隆·马斯克（Elon Musk）将未经许可的燃气轮机运到了他位于孟菲斯的 xAI 园区，该园区位于一个以黑人为主的社区，而该社区本已饱受高水平污染和环境破坏之苦。《华盛顿邮报》一篇广为流传（后遭质疑）的文章估计，单次 ChatGPT 搜索消耗的水量相当于一瓶水。随着数据中心的新增负荷导致公用事业公司预测需求将飙升，人们开始盯着自己的电费单。如果你相信各种经济媒体的说法，数据中心建设现在几乎支撑着整个美国经济。而这一切发生的同时，科技公司首席执行官们还在公开揣测人工智能将如何夺走所有人的工作，并取代人类的创造力本身。

People, naturally, started to ask questions.

人们自然开始提出质疑。