AI 是泡沫嗎?作者指出資金、技術與個人計劃須分開判斷
Is AI a bubble? Yes and no, and the difference matters to you
Saifuddin Ahmed 將 AI 泡沫之辯拆成資金、技術與個人計劃三個問題,認為估值可能下跌,但技術不會因此消失,不少以 AI 名義出售的事物則可能消失。他引用對大型雲端公司 2026 年資本開支的估算,約為 $660 billion 至 $800 billion。作者認為,較能承受市場下跌的計劃應立足於理解真實用戶、解決棘手問題、使用獲準使用的數據,以及設計、判斷和溝通等技能。
Everyone is asking whether AI is a bubble. Investors ask it about stocks. Founders ask it about funding. Students ask it about careers. They are asking three different questions with the same word, and that is why the debate goes in circles.
Let me separate them.
Question 1: Is the money a bubble?
The numbers are enormous. Analysts' estimates for what the largest cloud companies will spend on capital expenditure in 2026, mostly on AI data centres and chips, range from roughly $660 billion to $800 billion, depending on the tracker (Futurum Group estimate via banthebots.org, AL Capital Advisory). Critics point to rising debt, to companies funding each other's demand, and to revenues that have not yet caught up with the spending (Man Group).
Even people building AI have said it out loud. In August 2025, OpenAI's chief executive said investors as a whole were "overexcited" about AI. And in the same month, an MIT report on enterprise generative AI pilots found that most of them were not producing measurable returns yet.
So: could AI stock prices and private valuations fall hard? Yes. Very possibly. That is what bubbles look like from the inside.
Question 2: Is the technology a bubble?
Here history is useful. Between March 2000 and October 2002, the Nasdaq lost about three quarters of its value. Thousands of internet companies died.
The internet did not.
The fibre laid during the bubble was used for the next twenty years. The survivors of the crash became the largest companies on Earth. A bubble is often how a society overpays for infrastructure it will need anyway.
I see the same pattern every day in my own work. I use AI to draft, to review code, to translate between Arabic and English, to explain errors. Some of that value is real and permanent, and it does not disappear if a stock falls 50%.
So: will the technology vanish? No. But a lot of what is being sold as "AI" will.
Question 3: Is your plan a bubble?
This is the question that actually matters to most readers, and it is the one nobody asks.
Your plan is a bubble if it depends on any of these staying true forever:
- Cheap, subsidised AI. Many tools are priced below their real cost today. If your product only works at today's prices, it is fragile.
- One provider. If one company changes its terms, its model or its price, does your business survive?
- Hype as a moat. "We use AI" is not a product. What problem do you solve that people would pay for even if you never said the word "AI"?
- A skill that is just a prompt. If your entire career advantage can be copied by typing a better sentence, it will be.
And your plan is not a bubble if it rests on things that survive a crash: understanding real users, owning a hard problem, data you are allowed to use, and skills like design, judgement and communication.
When I built Rafu, a stock and credit book for small shops, I deliberately made it work with no internet and no cloud AI service. Not because AI is bad, but because a shop owner's records should not depend on anyone's funding round.
What to do if you are early in your career
- Learn AI tools deeply, but do not build your identity on one. Tools come and go. Judgement stays.
- Stay close to real problems. Bubbles burst far from customers first.
- Build things that work offline, cheaply, and without permission. They survive every cycle.
- Read the history of the 2000 crash. It is the best free course on what comes next.
On Ishraqa we published, in Arabic, a guide to preparing for AI in your job instead of panicking, and I wrote about what technology will take from you after 2030. This piece is the market side of the same question.
I'm Saifuddin Ahmed, founder of Ishraqa Daily, a community on Google Play, the App Store and the Microsoft Store that turns knowledge into impact: developer clubs where people build open-source projects with mentors, languages, reading and listening together, and scholarships and opportunities. I also built Rafu and founded OtwoX Digital Solutions. More of my work is at saifuddin.dev and on LinkedIn.
I drafted this piece with an AI writing assistant. That is part of the point: the tool is useful, and the judgement still has to be mine.
來源:DEV Community · dev.to