施耐德电气（Schneider Electric）认为其数据中心业务没有放缓迹象，但表示运营商必须对社区更加公开，说明其项目将消耗的电力、水和土地。作为电气基础设施、冷却和配电设备的供应商，施耐德对运营商的支出计划有着直观的了解，并且在持续建设中有着商业利益。施耐德创新与数据中心副总裁史蒂芬·卡利尼（Steven Carlini）表示，施耐德的项目储备没有显示出放缓的迹象。
Schneider Electric sees no slowdown in its datacenter business, but says operators must be more open with communities about the power, water, and land their projects will consume. As a supplier of electrical infrastructure, cooling, and power distribution equipment, Schneider has a direct view of operators' spending plans – and a commercial interest in continued construction. Steven Carlini, vice president of innovation and datacenter, said Schneider's pipeline showed no slowdown.
“而且施耐德与大型数据中心运营商有很多协议、供应链协议。”今年早些时候的报道表明，美国计划在2026年投产的产能中，实际在建的仅有一半，为新的“数据仓库”项目清理土地的活动也已陷入停滞。尽管如此，卡利尼表示美国公司正在以最快的速度建设，并将其与欧洲的规划官僚作风和其他障碍进行了对比。
"And Schneider has a lot of agreements, supply chain agreements with the larger datacenter operators." Reports earlier this year suggested only half the capacity planned for 2026 is actually under construction in the US, and land being cleared for new bit barn projects has plateaued. Carlini nevertheless said US companies were building as fast as they could, contrasting that with planning bureaucracy and other obstacles in Europe.
“但我认为将发生的最大变化是，运营商在试图获得许可和申请电力时的透明度水平，”卡利尼说。“美国传统的运营模式是，运营商与负责发放许可的地方司法管辖区进行谈判，并且几乎普遍涉及保密协议。因此，这些数据中心将要建造的地区的居民在施工开始之前对这些项目一无所知。”
"But the one thing that I think is going to be the biggest change is the level of transparency from the operators when they're trying to secure permits and when they're requesting power," Carlini said. "The way the model traditionally worked in the US is the operators would negotiate with the local jurisdictions that were responsible for issuing the permits, and there would be non-disclosures involved, almost universally.
在美国，针对数据中心建设的反对声音日益高涨。维权团体在去年底呼吁暂停建设新设施，纽约州则在今年夏天实施了叫停。新建筑的巨大规模也让许多人感到惊慌，例如，Meta的“许珀里翁”（Hyperion）设施计划达到5吉瓦的电力容量，并覆盖相当于曼哈顿很大一部分的面积。
So the residents where these datacenters were going to be built had no visibility into these projects until construction started." Opposition to datacenter construction has grown in the US. Campaign groups called for a moratorium on new facilities late last year, and New York state imposed a halt this summer.
智库皮尤研究中心（Pew Research Center）最近的一项民意调查发现，美国人对数据中心的警惕性越来越高，超过半数的人表示它们对环境有害、对家庭能源账单有害，并消耗用于冷却的水资源。
The sheer scale of the new builds has also alarmed many folk, with Meta's Hyperion facility, for example, planned to reach 5 GW of power capacity and cover an area equal to a significant chunk of Manhattan.
卡利尼表示，作为回应，建设和运营计算中心的 COMPANIES 需要更加坦诚地说明其计划将对当地产生什么影响。
A recent survey by think tank Pew Research Center found that Americans are increasingly wary of datacenters, with more than half saying they are bad for the environment, bad for household energy bills, and consume water resources for cooling.
“他们必须披露电力将从哪里来、他们将如何获得电力、他们的用水量、碳排放量、数据中心的效率、它的物理位置在哪里、设施的噪声水平、水源（这是一个大问题），以及他们将使用多少水，这是运营商传统上在公布时一直不够坦诚的另一个大问题。”
In response, companies building and operating compute sheds need to be more upfront about what effect their plans will have on the local area, Carlini said.
在弗吉尼亚州——全球的“数据中心之都”——亚马逊网络服务（AWS）的一位前水资源可持续发展项目经理正将他的前雇主告上法庭，指控其发布了关于该州用水量的虚假和误导性声明。
"They're going to have to disclose where the power is going to come from, how they're going to get the power, their water usage, their carbon emissions, the efficiency of the datacenter, where it is going to be located physically, the noise levels of the facility, the water sources, which is a big one, and how much water they're going to use is another big one that operators have traditionally not been very forthright in publishing."
卡利尼预计，许可要求将迫使企业进行更广泛的披露，紧随其后的将是设施运营后的报告义务。
In Virginia – the world's "datacenter capital" – an ex-water sustainability program manager at AWS is taking his former employer to court over claims it published false and misleading statements about water use in the state.
“然后他们还必须将其作为运营期间法规的一部分进行报告，因此情况不会是‘这是我们的预测，然后你获得了许可，就再也不需要报告了’，”他说。
Carlini expects permitting requirements to force greater disclosure, followed by reporting obligations once facilities are operating.
卡利尼补充说，数据中心所有者和运营商还必须赢得人心，业内一些人士在去年的一次活动中就已经认识到了这一点。
"And then they're going to have to report it as part of regulations during operations as well, so it's not going to be a case of here's our prediction, and then you get the permit and you never have to report again," he said.
“从公共关系的角度来看，这是运营商真正需要关注的另一个领域，因为在许多情况下，他们一直认为自己是在造福社区，”他说。“那些日子已经一去不复返了，因此从公关角度来看，运营商将不得不更加友好地与当地居民进行沟通。”
Datacenter owners and operators must win hearts and minds as well, Carlini added, a point already recognized by some in the industry at an event last year. "From a public relations perspective, that's another area where the operators really need to focus, because in many cases they've taken the position that they're benefiting the community," he said.
卡利尼告诉我们，人们使用Facebook或Gmail并不意味着他们会欢迎隔壁建一个数据中心。‘所以它必须更像是‘这就是我们将如何提供帮助’，”他说。
"Those days are gone, so the operators are going to have to, from a PR perspective, engage with the local residents more amicably." People using Facebook or Gmail doesn't mean they'll welcome a datacenter next door, Carlini told us.
运营商需要承认居民的关切并解释任何好处，例如税收减免或更低的成本。
"So it's going to have to be more like 'Here's how we're going to help,'" he said.
对于人工智能设施，卡利尼预计人们会更多地关注令牌处理性能和效率，而不仅仅是电力容量。
Operators would need to acknowledge residents' concerns and explain any benefits, such as tax breaks or lower costs.
“我认为我们将开始看到越来越多的关注点放在每瓦令牌数、每美元令牌数上，甚至数据中心可以按令牌来衡量规模，作为IT性能指标，而不仅仅是电力指标，”他告诉我们。“我认为该行业过去遭受过损失，因为数据中心一直仅按功耗来衡量规模，这让人很难了解它正在生产什么以及[它]正在增加什么价值。”
For AI facilities, Carlini expects more attention to token-processing performance and efficiency, rather than power capacity alone. "I think we're going to start to see more and more focus on tokens per watt, tokens per dollar, and even datacenters could be sized in tokens as a performance metric for the IT rather than just the power metric," he told us.
关于人工智能处理中令牌经济学的更多信息，读者可以参考我们今年早些时候关于这个话题的文章。®
"I think the industry has suffered in the past because datacenters have always just been sized in power consumption, and that gave little idea what it was producing and what value [it] was adding." For more on the economics of tokens in AI processing, readers can refer to our article on this topic from earlier this year. ®