Ulanqab's AI Data-Center Boom: 22 Facilities Under Construction, 30,000 Workers
A LatePost feature reports on the AI-era industrial boom in Ulanqab, China, where 22 data centers are under construction in the Chahar High-Tech Zone, employing nearly 30,000 blue-collar workers. The facilities house over 5,000 cabinets, each containing dozens of GPUs valued at over 500,000 yuan, with total investment exceeding 3 billion yuan and electricity consumption equivalent to 1.45 million households. Despite the massive construction, local life remains largely unchanged, with workers moving between sites every few months and small shops following the construction crews.
Coverage timeline
晚点 LatePost李梓楠
乌兰察布市察哈尔高新区巴音片区,卖方便面和饼干的小店开在集装箱里。店主 60 多岁,不会用豆包,知道工地里的人是搞大数据的。过去一年,他跟着开工的工地迁徙,从快手门口搬到中金,又搬到世纪互联,一共搬了 6 次。叫吊车搬一次,要花 300 元。 数据中心最忙的施工期通常只有三四个月。一座楼完工,几千名工人转去下一个工地,小店也得跟着走。等楼里的服务器运转起来,一位园区负责人算过,一栋楼值班和维护的人加起来不到 20 个。 小卖铺店主也感到周围逐渐冷清。去年园区几个数据中心集中启动建设,人流量极大,每天下班的时候周围都会堵车。为了赶工,工地设置了夜班,小卖铺也延长营业。10 月,这些数据中心陆续进入装修收尾阶段,现场工人明显少了,小店的生意也差了不少,“还能卖点烟和水,方便面一天都卖不掉一箱,一天只能赚 200 块钱”。 小店背后的园区占地 200 亩,投资额超过 30 亿,用电量与 145 万户家庭相当。里面有超过 5000 个两米多高的机柜,每个柜子放着数十张价值超 50 万元的 GPU ,它们像抽屉一样层层叠在里面。连接这些 GPU 的电线往往长数十公里,光纤的长度会超过一千公里。为了散去这些卡运算时产生的热量,还要安装近千台空调,用 2000 多吨铜片来散热。 从铜缆到高带宽内存,过去两年数据中心产业链每一个环节都在涨价,全球供应链都开足马力,共同造就我们这个时代最夸张的资本开支。 乌兰察布察哈尔高新区面积相当于 9 个上海迪斯尼乐园,仅这块地就有 22 个数据中心在建,近三万名日薪超过 300 元的蓝领工人在各个工地突击施工。他们每三个月就要建完一栋数据中心。这些工人每天工作 10 小时,能赚 200-400 元。 两个月前,高盛说乌兰察布已崛起为亚太地区规模最大、增长最快的 AI 算力集群。这座只有 150 万人,原本以土豆、火山和宇航员出名的北方边境小城,正在成为 AI 算力竞赛的前线。 但这里人们的生活,并没有被数据中心改变。 卡车运来了 GPU 和蜜雪冰城 从乌兰察布市中心出发,不超过 10 公里你就能看到数据中心。巨大的白色盒子从草原的平整地面拔地而起,两侧是柴油发电机,像监狱一样被栅栏和铁丝网围住,上面还通着高压电,每隔十米就有一个摄像头,角落立着一根冒白烟的柱子,这是数据中心的散热塔,不是烟囱。 它们是草原上最新、最气派且神秘的建筑。 出租车司机

Related stories
Data Labeling Industry Booms as AI Labs Seek Human Demonstration Data
A LatePost investigation details the growing large-model data-labeling industry, where workers in Chinese towns earn under 10,000 yuan monthly to fact-check, score, and write demonstrations for AI training. US labs like Anthropic collect data from real human work, while Chinese firms restructure data teams; the demand has spawned billion-dollar vendors like Surge AI and Mercor, and dozens of Chinese suppliers since early 2026.
1 reports · 1 sources

SemiAnalysis: Inner Mongolia is China's fastest-growing AI data center region
SemiAnalysis reports that Inner Mongolia has become China's fastest-growing AI data center region, driven by low construction costs and the absence of public backlash. The report highlights an 800MW wind-powered campus, 120MW built within a year, and a 5MW SST-native 800V DC pilot with a cooling design that omits central chillers.
1 reports · 1 sources
Big Tech AI data center buildout to continue despite lack of returns
The Verge reports that Amazon, Microsoft, Alphabet, and Meta plan at least $720 billion in capital expenditures this year, mostly for AI infrastructure, on top of over $1 trillion in prior commitments. Despite no closer to promised returns, the buildout will not slow on its own, requiring external intervention to stop it.
1 reports · 1 sources

Mature CNN/RNN and small LLMs deployed in industrial settings
LatePost reports that as frontier models dominate headlines, mature and cheaper CNN/RNN models and small LLMs are being deployed in real-world industrial settings such as logistics sorting centers and manufacturing lines. Examples include J&T Express using AI servers to monitor conveyor faults, an Apple supplier using vision models for hinge inspection, and judicial departments using a 27B Qwen model for archival processing. The article notes that these deployments often use GPUs from Denglin, which sold over 10,000 units in the first four months of the year.
1 reports · 1 sources

News updates every hour. Subscribe in your RSS reader to get new stories automatically.
Subscribe