Will AI Replace Programmers? What Industry Professionals Say

Short answer: no. But the job is changing, and getting your first job in it is harder than it was three years ago.

The strange thing about this question is that the people best placed to answer it do not agree. The CEO of one AI lab says software engineers have months left. The CEO of the company selling the chips says that view is backward. The man who built Linux says something different again.

So let's start with what they actually said, then look at what the job market did.

The CEOs do not agree with each other

Dario Amodei, Anthropic. The most aggressive prediction comes from him. At Davos in January 2026 he said we may be six to twelve months away from models doing most or all of what software engineers do end to end. He pointed to his own staff, some of whom no longer write code by hand and instead edit what the model produces. He also added a caveat that gets quoted less often: there is a lot of uncertainty, and it could take years.

Jensen Huang, Nvidia. He thinks the fear is pointed the wrong way. Speaking to founders at Y Combinator, he argued AI automates tasks, not whole jobs. His logic is simple economics: if programming gets cheaper, companies build more software, and building more software takes more engineers. He has publicly said he disagrees with almost everything Amodei says, though that remark was made about Amodei's wider warning on entry-level white-collar jobs, not about programming specifically.

Mustafa Suleyman, Microsoft AI. He walked back his own earlier warnings. In June 2026 he said AI will not take over white-collar jobs in the next 18 months, and will mostly handle boring tasks instead.

Sam Altman, OpenAI. He has softened too. He now says he does not expect the kind of jobs apocalypse some of his peers have described, though he still thinks coding is highly exposed to automation. His advice to students is to get good at using AI tools, the same way his generation got good at coding.

Keep one thing in mind while reading those four. Every one of them has money riding on AI, whether they sell the models or the chips that run them. Two predict upheaval, two predict calm, and they are all describing the same technology. That is a good reason not to lean too hard on any single prediction, including the confident ones.

‍

A study found most of the developer’s job isn't coding

One study is worth more than all of the predictions above.

First, how software actually gets built. It is a chain. Someone writes code. The code gets bundled into a commit. Commits get built up into a working project. The project gets tested, approved, and finally released to real users. A person has to check the work at every step before it moves to the next one.

Researchers tracked more than 100,000 GitHub developers along that whole chain, from the first line typed to the shipped release. Here is what coding agents did at each step:

• Commits: up 180%
• Finished projects: up 50%
• Actual releases: up 30%

The gain shrinks at every step. That is the finding.

If AI were really replacing developers, the number would stay high all the way down. It doesn't. It leaks away wherever a human still has to review the work and take responsibility for it.

Think of a factory. You buy a machine that cuts parts ten times faster. You now have a mountain of parts. But the number of finished products going out the door barely moves, because the same inspectors are still checking every one at the same speed. The fast machine didn't replace them. It just gave them a bigger pile to get through.

That is what happened to software. Writing code was never the slow part. A Microsoft study of around 6,000 developers found that even on the most generous estimate, close to 40% of a developer's time goes somewhere other than writing code, and most estimates put the coding share far lower than that. The rest goes on working out what to build, checking it, and being on the hook when it breaks.

How the job market actually shifted

Here the news is genuinely mixed, and you should be suspicious of anyone who tells you otherwise.

The bad part. Software development job postings in the US are still about 27.5% below where they were in February 2020. Overall job postings across all industries are roughly back to normal. Software did not recover with everything else.

The better part. Postings have been rising again. Indeed's Hiring Lab tracked a 15% rise in software development postings over the past year while overall postings fell 7%. Federal Reserve economists found that programmer employment is still growing, just about 3 percentage points a year slower than before ChatGPT.

The part nobody talks about. Look at who the rebound is for. Of the increase in software postings over the past year, 71% came from senior roles. And in the first quarter of 2026, senior positions made up 69.3% of all software development postings, the highest share of any occupation Indeed tracks. Across the whole US job market, only about 14% of postings are senior.

Entry-level postings, meanwhile, fell 7.5% over the year.

What about all the AI layoffs?

Companies blame AI when they talk to the press. They never blame AI when they talk to the government.

Here is how we know. In the US, a company has to warn the state government before it lays off a lot of people. It fills in a form, and that form is public. In March 2025, New York State added one question to it: did technology or automation cause these job cuts? If yes, tick the box and say which technology. In the next year, more than 160 companies filled in that form. Nobody ticked the box.

This is not proof. Nobody checks the answers, and there is no punishment for leaving the question blank. But a press release is written to impress investors. A government form is not. The same companies filled in both, and only one of them mentions AI.

Harvard Business Review asked 1,006 company bosses about their layoffs. Most of them had cut jobs because of what they believed AI would be able to do soon. Not because AI was already doing that work. They fired first and waited for the technology to catch up.

It often didn't.When  hiring managers were asked who had cut jobs after bringing in AI what happened next. More than three in ten had to hire people back into the same roles, or roles very close to them.

The real problem: nobody hires juniors anymore

Juniors learn by doing the simple work. You get the boring ticket, you write it badly, someone senior tells you why, and after a few hundred rounds of that you develop the instinct that later lets you spot a bad design before you can explain what is wrong with it.

Agents are excellent at exactly that tier of work.

So the step that produces senior engineers is the step being automated first, while 69% of software postings ask for people who have already climbed it. Forrester's forecast names junior roles as one of the categories under the most pressure, which fits.

The shift does not mean software development is disappearing. Instead, companies are increasingly looking for engineers who can work effectively with AI tools, review generated code, and manage more complex systems. This is driving demand for teams that combine traditional engineering skills with modern AI software development capabilities. 

‍

FAQ

Will AI replace programmers by 2030? 

The evidence points away from it. AI has sped up code writing, which is a minority of the job, while leaving planning and delivery mostly untouched. In a study of 100,000 developers, a 180% rise in commits produced only a 30% rise in releases. What is clearly changing is hiring speed and the number of entry-level roles.

Is it still worth learning to code in 2026? 

Yes, but the emphasis has moved. Knowing syntax is worth less. System design, writing clear specs, and reviewing code are worth more. Be realistic that the first job takes longer to find than it used to.

Are companies really firing developers because of AI? 

Some say so. The paperwork suggests otherwise. In the first year that New York State asked companies to declare it, more than 160 filed mass-layoff notices and none named automation as a cause. A survey of 1,006 executives found cuts were made in expectation of AI's impact rather than because of results, and over three in ten managers who cut roles later rehired them.

Which programming jobs are most at risk? 

Jobs defined narrowly around writing code to someone else's spec. The split between "programmer" and "software engineer" is decades old. AI is speeding it up, not creating it.

Is AI-generated code safe to ship? 

It depends on supervision. In the Stanford data, code written almost entirely by agents carried around nine times more security vulnerabilities per line than human-written code. Code written by a human and agent together was much safer. Review is the variable, not the model.