Five years ago I thought that as long as I worked hard, clocked enough overtime, and picked up enough skills, my life would be set.
Then reality hit. Last year a friend of mine—law school graduate, three years to finally pass the bar exam—watched his firm roll out AI contract review. The intern’s week of work, the AI did in two hours. His boss said directly: no new hires this year. He froze in the meeting room.
This is not an isolated case.
If you have been watching the news lately, you have seen a new term popping up: Agentic AI—sounds sci-fi, right? Let me rephrase it so it clicks:
In the past, AI was your chat assistant, your research helper. Now AI is a virtual employee that does the work for you—operates your computer, fills out forms, replies to emails, writes your code. It keeps working while you sleep.
Today I want to talk about three things:
- Why the path of living off a salary is having its floor pulled out
- If we really cannot go back, where should we put our money now
- What should our kids learn so they do not get eliminated
1. White Collars Are Falling First—The Information Asymmetry Premium Is Disappearing
Have you noticed that the people feeling the squeeze first are not factory line workers, but the white collars sitting in offices—travel agencies, real estate agents, paralegals, customer support, even coders.
What do they have in common? They are all bridges—the person connecting you to information, services, or products.
Travel agents compare prices for you. Real estate agents search listings. Lawyers research precedents. But now AI can scan the entire world’s airline tickets in seconds, then book your flight, fill in your itinerary, complete your visa form—all in one shot. The bridge in the middle is no longer needed.
Economists call this information asymmetry. Plainly put: you know too little, I know too much, so you pay me. But AI has made that kind of “knowing” worthless.
Globally, the services that charge for information aggregation, process outsourcing, professional document handling, and transaction matching add up to trillions of dollars. Now that cake is being eaten bite by bite by AI, and the pace is faster than anyone expected.
2. Agentic AI: Even Programmers Are Starting to Panic
In 2023 people were still saying “AI will replace repetitive work first.” But in reality, the high-paid white collars are feeling the chill first.
In January this year, Anthropic launched an AI agent platform. In plain terms: you give it a task, and it can break it down on its own, operate across software, find data, fill forms. You just say “organize this month’s contracts,” and it opens Excel, opens email, and handles everything for you.
The market reacted immediately. The value of many roles in law, accounting, and customer support is about to be wiped out.
Even more shocking, even programmers are starting to panic. Three or four years ago everyone said “go learn to code.” Now? Silicon Valley has slowed its developer hiring. Why? Because AI writes code without breaks, never calls in sick, never gets moody, and the output is not bad. What do you have to compete with?
In January this year, an Austrian developer built an open-source AI agent tool in his spare time, capable of connecting to computers and various apps, automating workflows. The news exploded across the industry—people had thought agentic AI like this would only mature by 2030, but the industry is now seriously discussing whether it can scale commercially by 2027 or 2028.
One side project pushed the entire industry’s timeline forward by a big step.
3. AI Is Pulling the Floor Out—Labor Income Is No Longer an Iron Rice Bowl
So here is the question—if the path of earning by working is getting narrower, what do we rely on?
People’s income is actually only two kinds:
- Labor income: you go to work, the boss pays you
- Capital income: your money works for you and earns more money
For decades the social logic was “study hard, work hard, your income will steadily grow.” But AI is breaking that assumption.
The CEO of Anthropic once said something I still remember:
“Powerful AI is like creating a nation of geniuses inside a data center. This nation has tens of millions of citizens, each one approaching top-scientist level. They work 24 hours without rest, never eat, never sleep, never strike. For us ordinary people to compete with this nation, how good do you think your odds are?”
So in the future we may face a flip:
In the past, labor income was 80 to 90 percent of income and capital income was a rounding error. In the future, capital income will grow more important, and may slowly become the lead.
This ratio will not flip overnight, but the direction is set, and the speed is faster than most people imagine.
4. So What Do You Buy? Don’t Bet on One Winner—Bet on What Every Winner Needs
When people talk about AI investing, they immediately think of Nvidia, Microsoft, OpenAI—but there is a huge risk here: you simply cannot predict who will win. The winners of three years ago may be left behind three years from now. OpenAI looks mighty now, but its cash burn rate is scary.
What to do? Change perspective—whichever AI company ends up winning, they will all need data centers, semiconductors, electricity, cooling systems, fiber.
Don’t bet on a single company. Bet on the things every winner has to use.
This is like the 1999 internet boom. Back then no one knew whether Yahoo or Amazon would survive, but everyone needed fiber. When the bubble burst many companies died, but the fiber remained. The survivors used that surplus infrastructure to scale at low cost.
AI will be the same—even if the bubble bursts, the data centers, chips, and power infrastructure will very likely remain.
There is also a change many people have not noticed: in the past the focus of AI investment was training—using massive data for the model to learn. But the real explosion ahead may be inference—the actual usage of AI. You chatting with GPT, asking AI to write code, asking AI to generate images—each call is inference.
And the rise of agentic AI will explode inference demand—because an AI agent does not just answer one question and stop. It breaks down tasks, opens tools, iterates, fixes errors. The compute volume produced by one AI agent per day is on a completely different order of magnitude from a traditional chatbot.
And inference does not necessarily need GPUs. GPUs are like heavy dump trucks on a construction site, powerful but fuel-hungry. For daily driving you use a sedan, fuel-efficient and flexible. Inference requires purpose-built chips—low power draw, high efficiency, low latency.
This is where the next opportunity hides. There is a technology called HBM (High Bandwidth Memory) you may have heard of; and there is something called HBF still in early R&D, designed to solve the memory problem in AI inference. Whoever can mass-produce it first may get the next ticket.
You do not need to master every detail right now, but you can remember this keyword and track its sample progress, customer list, and production capacity.
5. A Severely Underestimated Direction: Traditional Companies Using AI to Boost Efficiency
Beyond all this, there is one direction being severely underestimated—traditional companies using AI to dramatically boost efficiency.
Walmart’s stock has surged in recent years, and the reason is not just great retail execution—it is their large-scale deployment of AI in inventory management, logistics optimization, and customer support. Same with cars—the market used to value auto companies by auto-industry multiples, but now it is starting to value them as software and data companies.
Many Taiwanese manufacturers, if they can successfully integrate AI, could see their valuation logic rewritten too.
6. Your Portfolio Should Mimic the Stability of a Salary
Think about this—you get paid a salary every month, and this month your stocks also made NT$50,000. Are these two NT$50,000s the same?
Not at all.
The NT$50,000 from your salary will very likely still be there next month, and that certainty lets you spend, rent, plan long-term. But the NT$50,000 from stocks might be NT$80,000 next month, or minus NT$30,000—you do not know—so you cannot spend it, you can only save it.
This is the permanent income problem.
If more and more people shift their primary income from salary to investing, and investing is inherently unstable, what happens?
People stop spending → consumption shrinks → corporate revenue drops → stock prices fall → investment income becomes more unstable → a vicious cycle
That is why portfolio design in the AI era is more important than ever. You are not just chasing the highest return—you are trying to make your investment income mimic the permanence of a salary.
How? Just three words: diversify, diversify, diversify. Across assets, across sectors, across countries.
- For people in their 50s and 60s: lean more into high-dividend ETFs, bonds, REITs that produce stable cash flow, and let the rest chase growth
- For people in their 30s and 40s: balance growth and stability
- For people in their early 20s: the most important thing is not making money—it is using small amounts to try different ETFs, to understand how different assets behave in different market environments
ETFs are the best tool the ordinary person has in this era. Buy one and you automatically spread across dozens or hundreds of stocks. Buy five or six different types and your risk is even more diversified. NT$100,000 split into 10 parts across 10 different ETFs is fine. The point is not how much you make immediately—it is practicing the fundamentals.
One more reminder: three big variables stack up in 2026—US-China relations, the Federal Reserve chair transition, and the US midterm elections. These three are not isolated; they are deeply interconnected. The most dangerous strategy is putting all your money on one direction. The moresteady approach is broad diversification, staying flexible.
7. In the AI Era, What Should Our Kids Actually Learn?
No one can guarantee what your kid should learn. Anyone who says “learn X and you are definitely set” is either lying to you or lying to themselves. But a few points are reasonably solid:
First, the winner-takes-all effect will intensify
The AI era will lean further toward winner-takes-all. What you do matters less and less; being in the top 10 percent of your field matters more and more. AI squeezes the middle, and only the top performers, or people who know how to collaborate with AI, will have more opportunity.
Second, do not force your kid to imitate others
Is your kid obsessed with dinosaurs? Wonderful—let them become the person who knows dinosaurs the best. When people are genuinely interested in something, they build amazing depth on their own. Depth is 10,000 times more valuable than breadth.
Third, adaptability beats knowledge
The hottest skill today may be worthless in five years. The ability to learn fast and pivot fast is more valuable than any specific knowledge.
Fourth, resilience after failure
The more uncertainty the future holds, the higher the probability of failure. Failure itself is not the scary part—the scary part is how long it takes you to get back up.
Fifth, expose your kid to investing early
Not to teach them stock picking, but to use small sums to buy a few ETFs and let them watch what happens over five to ten years.
The hardest thing to learn in investing is called waiting. Waiting can only be learned through time. Start at 10, by 20 you have ten years of experience. That is worth more than any textbook.
8. The AI Era Could Be the Golden Window for Young Entrepreneurs
Another thing many have not thought about—the AI era could be the golden window for young entrepreneurs.
In the future, a single person can launch a company that previously required ten. AI helps you write code, design, customer support, ads—entrepreneurship costs drop off a cliff. Succeed, and you change your destiny; fail, and your loss is small.
This window will not stay open forever. Once most people becomeproficient, the edge disappears. Right now, it is still thefew.
9. You Cannot Choose Whether to Enter AI—Only Whether You Stand or Crouch
I am not going to lecture you. Let me share one sentence I really believe:
You cannot choose whether to enter this new world of AI. You can only choose whether to stand or crouch inside it.
Starting today, do one small thing—that is enough. Spend 15 minutes a day reading some financial news. You do not need to understand all of it. At first it is painful and normal—words you do not understand outnumber those you do by ten to one. But stick with it for a year and you will notice your speed picks up, and you start to have your own judgment.
Then write down your thoughts, no need to be long—just three to five hundred words. Read it once, write it once, and if you can discuss it with someone once, you have reviewed it three times.
Ten years from now looking back, you will find you have built a knowledge base no one can take from you. There is no shortcut, only time.
Disclaimer: This article reflects the author’s personal observations and market views, and does not constitute any investment advice. The AI industry and the policy environment carry high uncertainty. Any investment decisions should be based on your own financial situation, risk tolerance, and independent judgment. The author bears no responsibility for any investment gains or losses.
Comments