美国如何重塑制造业：对抗深圳的四大战略

✍ Zane Hengsperger🕐 2026-07-13📦 12.1 KB 🟢 已读 𝕏 文章列表

文章分析了美国制造业相对于深圳的衰退原因，提出了重振美国工业的系统性战略。作者指出关键在于文化重塑、建立工厂密度集群、将能源转化为武器以及实现供应链本土化。文章强调利用美国在软件和AI方面的技术优势，结合实体制造业的复兴，来建立不可替代的工业体系。

制造业宏观经济供应链深圳 AI 能源工业革命美国经济

# America can dominate Shenzhen. A call for factory abundance. **作者**: Zane Hengsperger **日期**: 2026-07-12T15:13:34.000Z **来源**: [https://x.com/zanehengsperger/status/2076324092286787941](https://x.com/zanehengsperger/status/2076324092286787941) --- ![Image](/doc/2950/image/img_585ba3298cd7.jpg) # Another industrial revolution is upon us. Shenzhen was a fishing village in 1980. Today it ships more electronics than the entire United States. You can go from CAD file to injection molded prototype in 48 hours there. In America that takes 6 weeks and 4 vendors. That is not because Chinese engineers are smarter. It is not because their labor is cheaper anymore either. Shenzhen wins because of density, speed, and a culture that treats building physical things as the highest-status work you can do. Every one of those is copyable. America copied nothing for 30 years and offshored instead. In 2000 the US was 25% of global manufacturing value added and China was 6%. By 2023 China was 29% and the US was 17%. That was a choice. We can make a different one. Here is a playbook, in order. And the order matters. ## 1. Fix the culture first I truly believe culture is everything. I think it is business, I think it is family, I think it is community, I think it is an industry, I think it is a nation. America has subpar culture. But it is fixable. This is the bottleneck under every other bottleneck. You can pour concrete and buy machines. You cannot buy a generation that wants to work in a factory. Only 6% of American high schoolers consider manufacturing a career path. For every 5 skilled tradespeople retiring, 2 are entering. The average welder is 55 years old. Meanwhile in Shenzhen, the guy running a CNC shop at 26 is the guy his friends want to be. We spent decades telling every smart kid the same thing: get out of the shop, get into the office. Learn to code. Manufacturing is what your grandfather did. The result is 570,000 open manufacturing positions and a trillion dollars of unfilled output on the table by 2030. The fix is status, not subsidies. Make the machinist the main character again. The factory worker should be sexy. What that looks like in practice: technicians walking factory floors and posting about it. Defense tech making hardware cool for the first time since Apollo. Trade schools that pay students to train on real jobs and place them before graduation. Twenty-two-year-olds seeing that the welder with pipe certs out-earns the marketing analyst. High schools with working shops instead of career pamphlets. I honestly think Instagram is good for this. I see so many cool young people working on machines, and it is becoming cool again. Culture moves before capital does. Nobody builds a factory in a country where nobody wants to work in one. Solve this and every number below gets easier. Skip it and none of the rest matters. ## 2. Build 100x more factories We should be aiming for factory abundance. Shenzhen's superpower is density. Thousands of factories within a 50-mile radius. Your PCB shop, your injection molder, your CNC house, your final assembly line, all neighbors. Iteration cycles measured in hours. America has the opposite. Our industrial base is scattered, aging, and shrinking. Six primary aluminum smelters, four partially or fully curtailed. 79 million tonnes of steel a year against China's 1.005 billion. Lead times of 8 to 30 weeks on basic metal. You do not out-compete density with a handful of mega-fabs and a press release. You out-compete it with volume. Hundreds of new mills, foundries, machine shops, fab shops, and assembly plants, clustered on purpose. Cluster them where the bones already exist: Detroit, Houston, Phoenix, the Carolinas. Cheap land, existing freight rail, workers with muscle memory. Co-locate the supply chain so a part never travels more than a day to its next operation. That is what Shenzhen actually is. Not one big factory. Ten thousand small ones that function as a single organism. And build them with energy designed in from day one. Aluminum and steel are energy monsters. On-site generation and next-gen nuclear are finally real. A factory with its own power is a factory China cannot underprice forever. ## 3. Turn energy into a weapon American electricity demand was flat for two decades, growing 0.1% a year from 2005 to 2019. Then AI showed up. US data center power demand is on track to more than double from 31 GW in 2025 to 66 GW by 2027. Data centers will pull 8.5% of peak summer demand by 2027, up from 4% two years earlier. In Virginia, they already consume more than 1 in 4 kilowatt-hours in the state. Texas grid load is growing 10% a year. A mill and a data center are bidding on the same electrons. Aluminum smelting is one of the most power-hungry processes on earth, and it is bidding against hyperscalers with trillion-dollar balance sheets who will pay whatever the market clears. On a legacy power contract, the mill loses that auction every single time. That is how smelters go dark in Missouri while server farms break ground down the road. The losing move is to complain about it. The winning move is to flip it. Energy-intensive industry has to stop being a grid customer and become a grid asset. Mills with generation on-site: gas now, next-gen nuclear and geothermal as they land. Mills as flexible load: run hard when power is cheap, shed 100 MW back to the grid when the data centers spike demand, and get paid for it. A smelter that can curtail on command is a virtual power plant that also happens to make metal. Co-locate mills and data centers around shared generation and shared interconnection, because the queue for a new grid hookup is now measured in years and a queue slot is worth more than the land under it. And the surge itself is the tailwind. AI is forcing America to build generation at a pace not seen in two generations. Every gigawatt built for compute is grid muscle heavy industry can lean on for the next 40 years. Cheapest firm power on earth wins heavy industry, full stop. The data center boom is dragging us toward it whether Washington plans for it or not. Build the mills next to it. ## 4. Anything you need, here I want to be able to need 8 parts from 8 different manufacturing processes and get them all on US soil. Shenzhen's real product is not cheap labor. It is proximity. Need a gearbox, a coating, a mold, a custom bracket at 4pm? Someone within an hour makes it, and it is on your dock by morning. That is the entire 48-hour iteration loop. It is not one supply chain. It is every supply chain, within reach. America's version does not need to be one hour. It needs to be one border. Anything a factory needs, sourced on US soil, in days. Ore to alloy, casting to coating, fastener to finished machine. That is the standard. We are nowhere near it. Today the honest test fails almost everywhere: can you build a complete product, an engine, a missile, a machine tool, without a single PO crossing the Pacific? China refines roughly 90% of global rare earths, and every F-35 carries 920 pounds of them. Our machine tools come from abroad. Our aluminum leans on imports. The supply chain has kill switches, and we do not control them. So map every critical input and close the gaps one company at a time. Rare earth refining. Forgings and castings. Specialty alloys. Machine tools. The unsexy middle of the stack that nobody tweets about and everything depends on. Then put a software layer over the whole thing, so a shop in Ohio can type "6061 sheet, anodized, 500 pieces, Thursday" and get it from 300 miles away instead of a container ship. The end state is a promise you can print on a wall: whatever you need to build, you can get it here. When that is true, reshoring stops being patriotic and starts being the obviously correct business decision. That is when the flywheel turns on its own. ## 5. Leverage the tech. This is our unfair advantage. Code is great. Code stuff. But the capacity comes first. Here is the part everyone gets backwards. China built capacity first and is layering technology on top. The US tried to skip capacity and own only the technology layer. That trade failed. But it left us holding something valuable: the best software, AI, and robotics talent on earth. The move is not software instead of factories. It is software inside factories, from day one. China passed us in robot density in 2023 and runs roughly 5x more industrial robots than we do. But their software stack is not better than ours. Ours is better than theirs. We just never pointed it at the floor. A greenfield American factory in 2026 should be AI-native the way a startup is cloud-native. Quoting in hours, not days. Real-time scheduling instead of whiteboards. Predictive maintenance instead of $2M/hour unplanned downtime. Cobot cells doing the repeatable work so the certified welder does the work robots cannot. Every repair, every torch pass, every override captured as data, because the 30-year TIG welder has 10,000 micro-decisions in his hands and he retires in 4 years. Run the floor, get the data. Data trains the software. Software runs the floor better. The loop compounds, and the companies that vertically integrate both sides win. A brand-new American plant with 2026 software beats a 2005 Chinese plant with 2015 software. We get to skip their legacy. That is the whole damn advantage of starting late. ## 6. Get the government pulling the same direction I think the US is good at this and it's mostly bipartisan, but gov needs to play a huge role. Shenzhen did not happen by accident. Beijing designated it a Special Economic Zone in 1980 and then got out of the way on everything except capital, land, and speed. The state cleared the path. The market ran through it. America's government mostly does the opposite. Permits that take years. Environmental review that outlasts the product cycle. SBA loans capped at $5 million when the machine costs $8 million. Financing built for real estate, not for CNC cells. What Washington should actually do, and only this: Speed. Industrial permits in 90 days, not 5 years. If China can permit a smelter faster than we can permit a parking lot, we lose on paperwork alone. Demand. The Pentagon is the best anchor customer on earth. Multi-year buys at volume, the way the recent munitions ramp-ups proved possible. The 2027 deadline to get Chinese rare earths out of US weapons systems is the right kind of forcing function. Set more deadlines like it. Capital. Loan guarantees for industrial equipment and facilities, underwritten by people who understand what a machine tool produces. The Made in America Loan Guarantee is a start. 10x it. Then stop. No picking winners, no industrial five-year plans, no bureaucrats designing factories. Clear the path, anchor the demand, guarantee the capital, get out of the way. ## The through-line Shenzhen took 45 years. We do not have 45 years, and we do not need them, because we are not starting from a fishing village. We are starting with the deepest capital markets, the best software talent, and the strongest demand signal (defense plus reshoring) in the world. Fix the culture so people want to build. Build the factories so there is somewhere to work. Own the power so nobody can out-bid you for electrons. Close the supply chain so anything you need is already here. Load it all with technology so it runs faster than anything in Guangdong. Point the government at speed, demand, and capital, and nothing else. Shenzhen is not magic. It is density, speed, and status. All three are copyable. All three are beatable. Go build factories. ## 相关链接 - [Zane Hengsperger](https://x.com/zanehengsperger) - [@zanehengsperger](https://x.com/zanehengsperger) - [464K](https://x.com/zanehengsperger/status/2076324092286787941/analytics) - [Upgrade to Premium](https://x.com/i/premium_sign_up) - [11:13 PM · Jul 12, 2026](https://x.com/zanehengsperger/status/2076324092286787941) - [464.4K Views](https://x.com/zanehengsperger/status/2076324092286787941/analytics) - [View quotes](https://x.com/zanehengsperger/status/2076324092286787941/quotes) --- *导出时间: 2026/7/13 15:52:08* --- ## 中文翻译 # 美国可以主导深圳。呼吁实现制造业的富足。 **作者**: Zane Hengsperger **日期**: 2026-07-12T15:13:34.000Z **来源**: [https://x.com/zanehengsperger/status/2076324092286787941](https://x.com/zanehengsperger/status/2076324092286787941) --- ![Image](/doc/2950/image/img_585ba3298cd7.jpg) # 另一场工业革命已至。 1980年的深圳还是个渔村。如今，它出口的电子产品比整个美国还多。在那里，你可以在48小时内完成从CAD设计文件到注塑原型。在美国，这需要6周时间和4家供应商。这并不是因为中国工程师更聪明。也不再是因为他们的劳动力更便宜。深圳胜出，靠的是密度、速度，以及一种将制造实物视为最高地位工作的文化。这些因素都是可以复制的。美国在过去30年里没有复制，反而选择了离岸外包。2000年，美国占全球制造业增加值的25%，中国占6%。到了2023年，中国占29%，美国占17%。这是一个选择。我们可以做出不同的选择。这是一份行动路线图，按顺序排列。顺序至关重要。 ## 1. 首先修复文化我真心相信文化就是一切。我认为它关乎商业、家庭、社区、行业，也关乎国家。美国的文化质量欠佳。但它是可以修复的。这是所有其他瓶颈之下的根本瓶颈。你可以浇筑混凝土、购买机器。但你买不到一代愿意在工厂工作的人。只有6%的美国高中生认为制造业是职业选择。每有5名技术工人退休，只有2人入行。普通焊工的年龄是55岁。而在深圳，一个26岁经营数控机床（CNC）车间的人，是他朋友们向往的对象。几十年来，我们告诉每个聪明的孩子同一件事：离开车间，坐进办公室。学会编程。制造业是你祖父那一辈才做的事。结果是，到2030年，我们将面临57万个制造业岗位空缺，以及高达一万亿美元的产能缺口。解决办法在于提升地位，而不是提供补贴。让机械师重新成为主角。让工厂工人变得令人向往。在实践中，这看起来是这样的：技术人员在车间巡视并发布相关内容。国防科技让自阿波罗计划以来硬件再次变得酷炫。职业学校让学生在真实岗位上培训并获得报酬，并在毕业前安排工作。22岁的年轻人看到拥有管道认证的焊工收入超过了营销分析师。高中配备运转的车间，而不是职业手册。老实说，我认为Instagram在这方面很有用。我看到很多很酷的年轻人都在操作机器，这正在再次变得流行。文化先行，资本随后。没有人会在一个没人想进工厂的国家建厂。解决这个问题，下面的每一个数据都会更容易。跳过这一步，其余的一切都无关紧要。 ## 2. 建设100倍数量的工厂我们的目标应该是制造业的富足。深圳的超能力是密度。在50英里半径内有数千家工厂。你的PCB供应商、注塑厂、CNC车间、最终装配线，全是邻居。迭代周期以小时计。美国恰恰相反。我们的工业基础分散、老化且正在萎缩。六家原铝冶炼厂，其中四家已部分或完全停产。年产7900万吨钢材，而中国是10.05亿吨。基础金属的交货周期为8到30周。你无法靠几个超级晶圆厂和一份新闻稿来击败密度。你必须靠数量来超越。数百座新建的轧钢厂、铸造厂、机加工车间、制造车间和装配厂，并有目的地聚集在一起。将它们聚集在骨架尚存的地方：底特律、休斯顿、凤凰城、卡罗莱纳。廉价土地、现成的货运铁路、拥有肌肉记忆的工人。让供应链就地协同，让零件到下一道工序的运输时间从未超过一天。这才是深圳的真面目。不是一个大工厂，而是一万个作为一个有机体运作的小工厂。并且，从第一天起就将能源考虑在内。铝和钢是能源巨兽。现场发电和下一代核能终于成为现实。拥有自备电力的工厂，是中国无法永远在价格上压倒的工厂。 ## 3. 把能源变成武器过去二十年，美国的电力需求一直持平，2005年至2019年间每年增长0.1%。然后AI出现了。美国数据中心的电力需求预计将从2025年的31吉瓦（GW）翻倍不止，到2027年达到66吉瓦。到2027年，数据中心将消耗夏季峰值需求的8.5%，高于两年前的4%。在弗吉尼亚州，它们已经消耗了该州超过四分之一的电量。德州的电网负荷每年增长10%。轧钢厂和数据中心在争夺同样的电子。铝冶炼是地球上最耗电的过程之一，它正与拥有万亿美元资产负债表的超大规模企业竞争，后者愿意支付市场出清的任何价格。在传统的电力合同下，轧钢厂每次都会输掉这场拍卖。这就是为什么密苏里州的冶炼厂熄灯的同时，服务器农场在隔壁破土动工。失败的做法是抱怨。成功的做法是反戈一击。能源密集型产业必须停止充当电网客户，转变为电网资产。自带发电能力的轧钢厂：现在用天然气，待下一代核能和地热落地后接入它们。把轧钢厂作为灵活负载：电力便宜时全力生产，当数据中心需求激增时向电网回供100兆瓦，并为此获得报酬。一个可以随时削减负荷的冶炼厂就是一个碰巧还能生产金属的虚拟电厂。围绕共享发电设施和共享并网点将轧钢厂与数据中心聚集在一起，因为现在新电网接入的排队时间以年计算，而一个排队名额比它下面的土地更值钱。而且，激增本身就是顺风。AI正迫使美国以两代人未曾见过的速度建设发电能力。为计算建设的每吉瓦产能，都是重工业在未来40年可以依托的电网肌肉。地球上最便宜的稳定电力赢得重工业，句号。无论华盛顿是否计划，数据中心的繁荣都在把我们拖向那个方向。把轧钢厂建在它旁边。 ## 4. 你需要的任何东西，这里都有我希望当我需要来自8种不同制造工艺的8个零件时，都能在美国本土获得它们。深圳的真正产品不是廉价劳动力。它是邻近性。下午4点需要一个齿轮箱、涂层、模具或定制支架？一小时内的某人就能做出来，第二天早上就出现在你的码头。这就是完整的48小时迭代循环。它不是一条供应链。它是触手可及的所有供应链。美国的版本不需要达到一小时的距离。它需要在一个国境内。工厂需要的任何东西，都在美国本土采购，数天内搞定。从矿石到合金，从铸造到涂层，从紧固件到成品机器。这才是标准。我们离这还很远。如今，诚实的测试几乎在任何地方都失败了：你能不能在不开出一张跨越太平洋的采购订单（PO）的情况下，制造出一台完整的产品、一个引擎、一枚导弹或一台机床？中国精炼了全球约90%的稀土，而每一架F-35战机都携带920磅稀土。我们的机床来自国外。我们的铝材依赖进口。供应链上有“终止开关”，而我们并不掌握它们。因此，要绘制出每一项关键投入的地图，并一家接一家公司地填补空白。稀土精炼。锻造与铸造。特种合金。机床。那些没人发推特谈论、但一切又都依赖于它的无趣的中间层。然后，在整个体系之上铺设一层软件，让俄亥俄州的一家车间可以输入“6061铝板，阳极氧化，500件，周四取货”，并从300英里外而不是集装箱船上收到货。最终状态是一个可以印在墙上的承诺：无论你需要构建什么，你都能在这里得到。当这成为现实时，回岸（reshoring）就不再是一种爱国行为，而开始成为显然正确的商业决策。那就是飞轮自动转动的时候。 ## 5. 利用科技。这是我们的不公平优势。代码很棒。去写代码。但产能优先。这是每个人都搞反了的部分。中国先建立了产能，然后在上面叠加技术。美国试图跳过产能，只拥有技术层。这笔交易失败了。但这让我们留下了一些有价值的东西：地球上最好的软件、AI和机器人技术人才。举措不是用软件代替工厂。而是从第一天起，就将软件嵌入工厂内部。 2023年，中国机器人的密度超过了我们，其工业机器人的运行数量大约是我们的5倍。但他们的软件栈并不比我们好。我们的比他们好。我们只是从未将它指向过车间地面。 2026年的美国新建工厂应该是AI原生的，就像初创公司是云原生的一样。报价以小时计，而不是天。实时调度，而不是白板。预测性维护，而不是每小时200万美元的意外停机。协作机器人（Cobot）单元负责重复性工作，让认证焊工去处理机器人做不到的工作。每一次维修，每一次焊枪送丝，每一次参数调整都被捕捉为数据，因为那位30年的TIG焊工手中掌握着1万个微决策，而他还有4年就要退休了。管理车间，获取数据。数据训练软件。软件更好地管理车间。这个循环会复利增长，垂直整合两端的公司会赢。一家拥有2026年软件的全新美国工厂，将击败一家拥有2015年软件的2005年中国工厂。我们可以跳过他们的历史包袱。这就是起步晚的全部优势所在。 ## 6. 让政府朝同一方向用力我认为美国在这方面做得很好，而且这主要是两党一致的，但政府需要发挥巨大作用。深圳并非偶然。北京在1980年将其指定为经济特区，然后除了资本、土地和速度之外，其余方面都退居一旁。国家扫清了道路。市场在其中奔跑。美国政府的做法大多相反。许可证耗时数年。环境审查比产品生命周期还长。小企业管理局（SBA）贷款上限为500万美元，而机器成本为800万美元。为房地产而非数控单元设计的融资。华盛顿实际上应该做的，仅限于此：速度。工业许可在90天内完成，而不是5年。如果中国许可冶炼厂的速度比我们批准停车场的速度还快，我们仅仅因为文书工作就会输掉。需求。五角大楼是世界上最好的锚定客户。按批量进行多年采购，就像最近的弹药增产所证明的那样。2027年将中国稀土剔除出美国武器系统的最后期限是那种正确的强制功能。设定更多这样的最后期限。资本。为工业设备和设施提供贷款担保，由懂得机床生产什么的人来担保。《美国制造贷款担保计划》是个开始。把它放大10倍。然后停止。不要挑选赢家，不要搞五年计划，不要让官僚设计工厂。扫清道路，锚定需求，担保资本，然后退居一旁。 ## 贯穿始终的主线深圳用了45年。我们没有45年，我们也不需要45年，因为我们不是从渔村开始。我们起步于世界上最深的资本市场、最好的软件人才和最强劲的需求信号（国防加回岸）。修复文化，让人们愿意建设。建设工厂，让人们有工作可去。掌控能源，这样就没有人能在电子上出价高于你。闭合供应链，让你需要的任何东西都已在此处。给所有环节加载技术，使其运行速度超过广东的任何东西。指引政府关注速度、需求和资本，别无其他。深圳不是魔法。它是密度、速度和地位。这三者都是可复制的。这三者都是可战胜的。去建工厂吧。

🔗 相关文档