Is the AI Bubble Real?

The truly dangerous thing about the AI bubble isn't the technology.

It's that the entire world is front-loading financing for "decades of future intelligence demand" — all at once.

During the mobile internet era, people burned cash, sure. But they found revenue fast. Food delivery had customers. Ride-hailing had riders. E-commerce had buyers. Short videos had viewers. Ads had advertisers. Consumer-facing sectors — food, clothing, housing, transport, communication, entertainment, shopping — were all low-hanging fruit. The business loop closed quickly.

But AI is different. To this day, most of what's genuinely deployed at scale is still: writing weekly reports, making slides, generating images, customer service bots, coding assistants. Valuable? Yes. That's not the problem.

The real problem is this: the capital markets are already betting at the scale of "everyone consuming intelligence all the time." GPUs bought first. Data centers built first. Debt taken on first. Valuations pumped first. Pension funds entered first. The world is building "intelligence power plants" at unprecedented speed.

But here's the question: who, exactly, is going to consume intelligence the way we consume electricity today — continuously, at scale?

The biggest gamble in AI right now isn't whether models will get smarter. It's whether the explosion of B2B vertical applications can outpace the depletion of funding, GPU depreciation, data center debt, and the capital market's patience.

If Agentic AI genuinely penetrates core enterprise workflows — turning productivity gains into real profits — then a lot of today's crazed investments will be vindicated by history. But if the growth of real demand moves slower than the pace the capital markets have already priced in, then a lot of what today represents "the future" of AI may end up as: piles of power-hungry GPUs generating insufficient cash flow.

The railroad changed the world. Railroad stocks still crashed. The internet changed the world. Dot-coms still littered the battlefield. AI will probably change the world too. But a technological revolution being real has never meant a bubble doesn't exist.

发布者

立委

立委博士,多模态大模型应用咨询师。出门问问大模型团队前工程副总裁,聚焦大模型及其AIGC应用。Netbase前首席科学家10年,期间指挥研发了18种语言的理解和应用系统,鲁棒、线速,scale up to 社会媒体大数据,语义落地到舆情挖掘产品,成为美国NLP工业落地的领跑者。Cymfony前研发副总八年,曾荣获第一届问答系统第一名(TREC-8 QA Track),并赢得17个小企业创新研究的信息抽取项目(PI for 17 SBIRs)。

发表回复

您的邮箱地址不会被公开。 必填项已用 * 标注

这个站点使用 Akismet 来减少垃圾评论。了解你的评论数据如何被处理