Is it a bad idea to start a tech career right now because of AI?

The headlines make it sound like you would be walking into a building that is already on fire. The truth is more specific than that, and once you see the specifics the decision gets a lot clearer.

New to cloud? CAMPUX is a free, build-first course. Start here →

No, starting a tech career is not a bad idea because of AI — but you have to start in the right corner, because AI eats the typing, not the judgment, and the two are not evenly spread across tech. The generic entry path did narrow, honestly, and pretending otherwise helps nobody. What follows is where the risk actually landed, and where the work moved to.

The fear, said out loud

Here is the worry in plain words: you spend six months and real money learning a skill, and by the time you are ready to apply, a model does the job for free. That is not a stupid fear. It is the rational response to a year of layoff headlines, screenshots of AI writing whole apps, and executives saying out loud that they plan to hire fewer juniors. If you felt nothing reading that, you would not be paying attention.

So I am not going to talk you out of the fear. I am going to narrow it. The question that matters is not "is AI coming for tech." It is "which parts, and is the part I would start in getting bigger or smaller." Those have different answers, and averaging them into one number is how people talk themselves into the wrong decision.

What AI actually replaced in the last 18 months

Strip away the headlines and look at what changed on the ground. AI got very good at producing a first draft of anything with a clear specification: a function, a unit test, a config file, a page of marketing copy, a SQL query. Where the answer was mostly typing and the shape was known, output per person went up sharply, and teams stopped needing a junior whose whole value was fast, correct typing.

What did not get replaced is the part where someone decides what should be built, notices the draft is subtly wrong, and owns the consequences when it ships. A model will happily generate a deployment that opens a security hole, and it will do it confidently. Someone with judgment has to catch that. The 18-month lesson is narrow and worth memorizing: the closer a task sat to pure text generation, the harder AI hit it; the closer it sat to running a live system that other people depend on, the less it moved.

AI eats the typing. It does not eat the accountability for what the typing produced.

The honest part nobody admits: yes, generic entry-level got harder

I am not going to gaslight you. The old easy path — do a general coding bootcamp, put "full stack" on a résumé, apply to five hundred junior openings — got noticeably worse, and it was already crowded before AI. The rung that AI hit hardest is exactly the rung a lot of career-changers were told to climb. If someone tells you nothing changed, they are selling something.

Two things are true at once, and you have to hold both. The generic door narrowed. And the total amount of infrastructure that has to be built, run, secured, and paid for went up, because every one of those AI systems has to run somewhere. Both facts are real. The mistake is letting the first one hide the second. We wrote a longer, numbers-first version of this specifically about the entry rung: is AI reducing entry-level cloud jobs.

Why "build the infrastructure AI runs on" is the safest corner to start in

Every model you have read about runs on servers, storage, networks, identity systems, and monitoring — a mountain of cloud infrastructure that someone has to stand up and keep alive. When a company decides to use more AI, that is not a decision that removes work from the infrastructure team. It adds work: more compute to provision, more data to move, more security surface to watch, more bills to keep from exploding. AI is a customer of cloud, and you would be starting on the supply side of that transaction.

That is the whole argument, and it is structural rather than a slogan. In most of tech, AI is a substitute for some of the labor. In cloud and infrastructure, AI is a source of demand for the labor. It writes a config faster, sure, but it also creates ten more systems that need configs, and it cannot be the one accountable for whether they stay up at three in the morning. If you want the full breakdown of which roles that logic protects, we sorted it here: which tech jobs will survive AI.

A 90-day starting plan that assumes AI exists

This plan is not "learn to type code faster than a model," because you will lose that race. It assumes AI is on your desk from day one and your value is judgment about running things. Roughly:

• Weeks 1–3: the ground. Learn what cloud actually is — compute, storage, networking, identity — one concept at a time. Use AI to explain things you do not get, but write the summaries yourself so the understanding is yours, not the model's.
• Weeks 4–8: build one real thing. Stand up a small but complete piece of infrastructure in a real cloud account, in code, not clicked by hand. Let AI draft the config, then break it on purpose and fix it, because the fixing is the skill nobody can hand you.
• Weeks 9–12: prove and package. Add a second build, put both in public repositories with honest write-ups of what broke and how you diagnosed it, and get a foundational certification to clear the résumé filter. Then start applying to cloud support and junior operations roles, not generic "developer" postings.

Notice what the plan optimizes for: evidence that you can run a system and reason about it when it misbehaves. That is the thing AI does not supply, so that is the thing you sell.

How to tell in 3 weeks if this is for you

You do not need to bet six months to find out. Give it three weeks of real contact and watch your own reaction. When a deployment fails with a wall of red text, do you feel a pull to figure out why, or a wall of dread and a wish to close the laptop? The people who last in this work feel a version of curiosity there — annoyed curiosity, but curiosity. If after three honest weeks the puzzle bores you rather than nagging at you, that is useful information, and it is far cheaper to learn now than after a career change.

The way to run that test is to do the first class, not to read more threads of strangers arguing about the job market. Class One is free and takes about twenty minutes. If you want the fuller plan for switching careers with a real life to work around, we wrote that too: a career change into cloud, for people with bills. And if your fear is more about age than AI, that one has its own honest answer: is it too late to get into tech.

Common questions