Business · 2025-11-27
Tech Skeptic & Coffee Drinker (科技怀疑论者兼咖啡饮用者)

Is the AI Boom Just a Giant Shell Game? Let’s Expose the Playbook

人工智能繁荣只是场大型空壳游戏?让我们揭开这本操作手册

Is the AI Boom Just a Giant Shell Game? Let’s Expose the Playbook
www.derekthompson.org

每场关于人工智能的讨论都像坏掉的唱片——同样的数据、同样的研究、同样的回音室。科技巨头在烧钱?在赚钱?METR说生产力上升?下降?全都是从十几档播客和一份高盛报告里翻来覆去的老调重弹。

我们根本不是在辩论人工智能——我们只是在背台词。当资本支出达到GDP的1.4%,初创公司产品为零却能融资20亿美元,英伟达的「投资」神奇地变成了所有人的收入时,你不需要数据就能闻到泡沫的味道——你只需要还活着。

评论 (8)
Oracle Investor with Sleep Issues (睡不好的甲骨文投资者)
Oracle’s $100B+ debt and 14% margins on GPU hosting aren’t a red flag — they’re a five-alarm fire. You’re building trillion-dollar infrastructure on sand, betting the entire demand curve stays vertical. If AI growth slows even slightly, those $455B ‘firm commitments’ become toilet paper.

甲骨文公司超过千亿的债务和14%的GPU托管利润率不是红旗警告——这是五级火警。你们正把万亿美元级的基础设施建在沙地上,赌整个需求曲线永远垂直上升。一旦AI增长稍有放缓,那4550亿美元的「确定订单」就会变成厕纸。

Cloud Engineer Who's Seen Three Bubbles (经历过三次泡沫的云工程师)
Re: ‘Toilet paper’ — not quite. Many of those contracts are multi-year with penalties for early exit. But you’re right: when the AI music stops, the Oracle with the most debt gets left holding the GPUs.

回应「厕纸」一说——不完全对。很多合同是多年期且提前解约需支付罚金。但你说得对:当AI这阵音乐停下时,负债最高的甲骨文将只能抱着一堆GPU发呆。

AGI True Believer in SF (旧金山的通用人工智能信徒)
All big leaps look like bubbles at first. Railroads, internet, electricity — people called those speculative too. This isn’t a bubble. It’s a phase transition. The productivity data will catch up once the tools mature. Wait until AI can debug itself.

所有重大突破起初都像泡沫。铁路、互联网、电力——当年也都被说成投机。这不是泡沫,而是相变。等工具成熟,生产力数据自然会跟上。等AI能自己调试代码时再看。

Former Dot-com Survivor (前互联网泡沫幸存者)
‘Phase transition’? That’s what we said in ‘98. The ads bought with VC money paid for the revenue growth that justified more VC money. Sound familiar? This isn’t physics. It’s Ponzi-nomics.

「相变」?我们1998年就这么说。风投资金买的广告带来了收入增长,而收入增长又证明了风投的合理性。听起来耳熟吗?这可不是物理学,这是庞氏经济学。

AI Startup Founder with Secrecy Issues (有保密问题的AI初创创始人)
We raised $2B pre-product because trust > slides. The best teams don’t need to explain — they attract capital by vision alone. That’s how innovation happens.

我们在产品发布前就融资20亿美元,因为信任胜过PPT。顶尖团队无需解释——仅凭愿景就能吸引资本。创新就是这么发生的。

Skeptical Data Scientist (怀疑型数据科学家)
‘Vision’ doesn’t run inference. You can’t train a model on charisma. Show me the actual ROI, not the storytelling ROI.

「愿景」可没法跑推理。你不能用个人魅力训练模型。给我看真正的投资回报率,而不是故事回报率。

Macro Policy Wonk (宏观政策书呆子)
The scariest part? SPVs are moving AI risk off tech balance sheets and onto pension funds. If this bursts, it’s not just VCs losing money — it’s teachers and cops losing retirement savings.

最可怕的是?特殊目的实体正把AI风险从科技公司资产负债表转移到养老基金上。如果泡沫破裂,赔钱的不只是风投——而是教师和警察的退休金。

Retired High School Teacher (退休高中教师)
So my pension is funding AI startups that won’t tell anyone what they’re building? Wonderful. Let me update my résumé — Uber Eats might be hiring.

所以我的养老金正在资助那些连自己在做什么都不肯说的AI初创公司?太棒了。让我更新下简历——说不定优步外卖正在招人。