Will AI Replace Software Developers? What the Data Shows in 2026
Published October 19, 2026 · Updated November 22, 2026
Based on research, official statistics and surveys.
· Magyar változat
Research so far shows no broad, AI-driven unemployment among software developers. Entry-level developers are having a harder time, though: a Stanford analysis published in August 2026 finds that employment of 22–25-year-olds in the most AI-exposed occupations is 19% below the level implied by trends in less-exposed groups. The gap comes mainly from fewer new hires.
Demand for experienced developers is already picking up in US job postings; Europe has not yet seen a similar recovery. PwC’s analysis of more than a billion job postings puts the wage premium for jobs requiring AI skills at 62%, which does not mean that learning one AI tool brings a raise of the same size. In Hungary, official data show 62% fewer unfilled IT vacancies than in 2022, and companies use AI at half the EU rate.
AI is already taking over some development tasks. Whether that means fewer developers overall depends partly on how much new work cheaper software development makes worth doing.
Ever since ChatGPT launched in November 2022, people have been asking how long the world will still need programmers. Some expect the profession to disappear within a few years; others think AI’s impact is wildly overstated. The research paints a more nuanced picture.
This article draws on payroll data, controlled experiments with developers, analyses of job postings and official statistics to show who is most affected and what employers are looking for right now, with a closer look at Hungary.
1. What do job postings show? The US is recovering, Europe is not yet
Indeed’s software development postings index tracks how the number of job ads has changed relative to February 2020. It is a good gauge of demand, though postings are not the same as actual hires.
Monthly average. Source: Indeed Hiring Lab / FRED, October 9, 2026. Dashed line: February 2020 level.
| Country | 2022 peak | October 2026 | Trend |
|---|---|---|---|
| US | 233.8 | 78.1 | Rising since the May 2025 low of 61.1 |
| United Kingdom | 181.6 | 59.6 | Broadly flat |
| France | 141.0 | 51.7 | Still low |
| Germany | 169.0 | 47.6 | Lowest level since 2023 |
Source: Indeed / FRED, US, German, UK and French series. Values as of October 9, 2026.
The decline began in early 2022, months before ChatGPT was released. The timing alone shows that AI cannot explain it by itself: interest-rate hikes and the unwinding of headcount that had ballooned during the pandemic played a large part.
In the US, postings have been rising since mid-2025. According to an Indeed Hiring Lab analysis from July 2026, software development postings have grown by almost 15% since Claude Code was released in February 2025, while total job postings fell by 7%. Senior roles accounted for 71% of the growth, and roles with “AI” in the title for 37%; the two groups partly overlap. (Hiring Lab)
Europe, and Germany in particular, has not seen a similar rebound. Weaker economic growth and the troubles of German industry are likely factors. The gap between US and European data does not, on its own, tell us how large AI’s role is.
For job seekers, country and seniority matter a great deal. A stronger US market does not mean a graduate in Budapest or Berlin will find work more easily.
The longer view: 25 years of data
The Indeed index only starts in 2020. Official vacancy statistics go back further: the US Bureau of Labor Statistics (BLS) has measured the share of unfilled posts in the information sector since 2001, and Eurostat has done so for Hungary since 2008. Note that this sector covers telecoms and publishing as well as software.
US: BLS JOLTS; Hungary, Germany, EU: Eurostat jvs_q_nace2. Rate = vacancies ÷ (occupied + vacant posts). EU average shown dashed.
Seen this way, 2022 was an outlier: the US (6.8%), Germany (5.6%) and the EU (4.2%) all hit their highest rates that year, and all have fallen since. Hungary’s 1.8% in 2025 is roughly back at its 2012–2013 level, and Germany’s 2.7% is the lowest since its series began in 2011. The US rate (4.0%) remains above most years between 2001 and 2017. The 2008–2009 financial crisis caused an even steeper fall in the US (from 4.6% in 2007 to 1.8% in 2009), long before generative AI.
2. Why is it harder to land a first developer job?
Several studies find that in occupations more exposed to AI, it is mainly young and less experienced workers whose job prospects have worsened. The studies use different data and methods, so their percentages should not be compared directly.
Stanford Digital Economy Lab: payroll data, August 2026
Using ADP payroll records, the study finds that employment of 22–25-year-olds in the most AI-exposed occupations is 19% lower than trends in less-exposed groups would predict. No comparable gap appears for more experienced workers. The difference comes mainly from fewer new hires, and it shows up in occupations where AI tends to substitute for, rather than augment, human work. The estimate has grown with each update: 13% in August 2025, 16% in November 2025 and 19% in August 2026. The authors stress that the result is descriptive and does not by itself establish causation. (Stanford)
Harvard: about 65 million résumés and 281,000 firms
At the 10,433 AI-adopting firms in the sample, junior employment fell by roughly 9% over six quarters compared with non-adopting firms. No similar change was found for senior staff. The authors describe this as seniority-biased technological change. (Harvard)
Anthropic: labor market analysis, March 2026
The analysis ranks programmers among the occupations most exposed to AI: 75% of their tasks could be affected. That does not mean 75% of the work is already automated. It found no rise in unemployment, but the job-finding rate of 22–25-year-olds in affected occupations was about 14% lower. (Anthropic)
Federal Reserve Bank of Dallas, January 2026
The job-finding rate of 20–24-year-olds in the most AI-exposed occupations has fallen by more than 3 percentage points since its 2023 peak. The estimated effect on the national unemployment rate is far smaller: at most 0.1 percentage points. (Dallas Fed)
Relative employment gap for 22–25-year-olds in the most AI-exposed occupations, across three versions of the study. Source: Stanford Digital Economy Lab.
Juniors are often given well-scoped pieces of work: writing repetitive code, simple bug fixes, tests. AI tools are already useful for part of that. If a team can get the work done with its current headcount, the company may postpone hiring another junior. According to Indeed, 69% of software development postings in the first quarter of 2026 were for senior roles.
Our own sample, for reference only: 8.7% of Dev Radar listings that state a level target juniors or interns and 83% senior roles or above; in Hungarian listings the junior share is 17.2%. “Senior” appears in titles more often than “junior”, so treat this as a lower bound. Junior jobs
In Hungarian listings the share of entry-level roles is higher: 17.2% of the 290 listings that state a level are junior positions. There is a caveat: “senior” appears in job titles more often than “junior”, so these figures may understate entry-level opportunities.
Current junior roles: devradar.hu/en/jobs?seniority=junior
If fewer newcomers get a start for a sustained period, the pipeline of experienced engineers may narrow later on. That does not mean the market will automatically correct itself within a few years.
3. Does AI actually make development faster?
There is no shortage of vendor benchmarks and personal anecdotes about AI speed-ups. Studies that directly observe developers at work find that the result depends heavily on the task and the environment.
METR, July 2025
In a randomized controlled trial, experienced open-source developers worked on their own, familiar codebases. With AI they needed on average 19% more time to finish their tasks. Beforehand they had expected a 24% speed-up, and afterwards they still believed AI had made them faster. (METR)
Positive: faster with AI; negative: slower. Source: METR.
Even after finishing, participants thought AI had made them 20% faster, when in fact they were 19% slower. Their own perception and the measurement were 39 percentage points apart.
METR, February 2026
A follow-up measured shorter completion times: 18% faster for returning participants and 4% for newly recruited ones. METR nonetheless rated the result as weak evidence. A growing number of developers refused to work without AI, which may have skewed who took part. (METR)
Google DORA 2025
90% of the roughly 5,000 respondents used AI. The survey found a positive relationship between AI use and delivery throughput, but a negative one with stability. According to the report, how a team already works matters a lot: AI can help well-run processes and amplify the problems of disorganized ones. (DORA)
Stack Overflow, 2025 and 2026
In the 2025 survey of 49,000 developers, 84% were using or planning to use AI. 33% trusted the accuracy of its output and 46% did not. The most common frustration, cited by 66%, was answers that are almost right but need fixing, and 45% said debugging AI-generated code takes more time. In the 2026 survey, nearly 70% of respondents also used AI agents to help solve problems, but only 6.6% would trust AI with important work decisions. In 2025, 15% saw AI as a threat to their current job and 64% did not. (Stack Overflow)
Stack Overflow Developer Survey 2025, about 49,000 respondents. Source: Stack Overflow.
None of this yields a single speed-up figure that applies to every developer. When learning a new technology, prototyping or doing repetitive work, AI can save time. In complex existing systems, understanding, checking and fixing its suggestions can eat up that advantage.
So it is worth measuring the total time spent on a task, including testing and later fixes.
4. What does it mean when AI writes 75% of the code?
In April 2026 Sundar Pichai said that 75% of Google’s new code is generated by AI and then reviewed and approved by engineers. The figure he gave was more than a quarter at the end of 2024 and close to half in autumn 2025. (Google)
Company statements; the 2024 and 2025 values are rounded (“more than a quarter”, “close to half”). Source: Google.
Anthropic’s analysis classified 79% of conversations with Claude Code as automation, where the AI carries out the task, compared with 49% in the standard chat interface. These are the providers’ own figures and do not measure how automated developer work is as a whole. (Anthropic)
Why might experienced developers still be needed at these levels?
Much of software development does not directly produce new lines of code: clarifying requirements, designing systems, reviewing other people’s code, tracking down production bugs or organizing operations. The volume of AI-generated code says nothing about how much time a team spends on these.
Cheaper development can also make projects viable that previously were not worth the cost. This potential rise in demand is often explained with the Jevons paradox. There is no guarantee, however, that new work will fully offset the work automation removes.
Meanwhile, the range of tasks AI can complete on its own keeps growing. METR tracks how long a task a model can complete with a 50% success rate, measured by how long the same task would take a human. That length roughly doubled every seven months between 2019 and 2024, and has grown even faster since 2023. By 2026 the strongest models handle multi-hour tasks at that success rate. The estimate carries substantial uncertainty, about a factor of two. (METR)
A 50% success rate is still not enough for a company to hand such work to a model without review.
5. How much of the layoffs is AI?
Layoffs.fyi records the following layoffs at technology companies:
Source: layoffs.fyi, October 19, 2026. Other trackers follow different companies and announcements, so other figures appear in the press. The data are not limited to developers.
In 2026 several large companies cited AI when cutting staff or restructuring. Block laid off about 4,000 employees in February, roughly 40% of its workforce. Amazon eliminated 14,000 corporate roles in October 2025 and another 16,000 in January 2026. Its CEO had written in June 2025 that AI would likely reduce the company’s corporate workforce over the next few years. (Amazon)
Microsoft, by contrast, said its July 2026 cut of about 4,800 roles was not about replacing those jobs with AI. (TechCrunch roundup)
Company announcements make it hard to tell how many jobs were lost directly because of AI. A layoff can combine cost cutting, earlier over-hiring, business restructuring and automation.
Analyses by the Yale Budget Lab and the Dallas Fed found no significant, broad change in the occupational mix or in national unemployment linked to AI adoption through early 2026. Young workers and some exposed occupations may still have been hit.
6. What is happening to developer salaries?
PwC’s 2026 analysis of more than a billion job postings found a 62% wage premium for jobs requiring AI skills, up from 57% in 2025. Postings asking for AI skills grew by 69%, against 9% for all postings. (PwC AI Jobs Barometer 2026)
More than one billion job postings, global data. Source: PwC Global AI Jobs Barometer 2026.
Our own sample, for reference only: among non-US Dev Radar listings with a salary, AI roles have a median of about €11,246 gross per month vs €8,388 overall (851 AI listings). AI jobs · salary statistics
The US Bureau of Labor Statistics (BLS) projects a 7% decline in the “computer programmer” occupation between 2025 and 2035, citing automation and AI among the reasons. That is one specific US occupational category, not a forecast for all software developers. (BLS)
Projected US employment change. Source: BLS Software Developers, Computer Programmers.
7. Which skills are employers asking for?
In the World Economic Forum’s 2025 survey, employers expect 39% of workers’ core skills to change or become outdated by 2030. AI and big data top the list of fastest-growing skills, followed by networks and cybersecurity and technological literacy. The same employers expect 170 million jobs to be created and 92 million displaced by 2030, a net gain of 78 million. (WEF)
Share of employers in Hungary expecting increased use of each skill. Source: WEF Future of Jobs Report 2025, Hungary economy profile.
In Stack Overflow’s 2025 survey, 66% of respondents used JavaScript, 62% HTML and CSS, 59% SQL and 58% Python. Python use rose by 7 percentage points in a year. Building and running AI systems still relies on these fundamentals. (Stack Overflow)
Stack Overflow Developer Survey 2025, 31,771 respondents; multiple answers allowed. Source: Stack Overflow.
Our own sample, for reference only: 14% of Dev Radar IT listings for non-AI roles ask for AI coding tools (e.g. Copilot, Claude Code); our most requested technologies are Python, AWS, SQL, Azure, AI/ML. Tech trends
8. Is there still a developer shortage in Hungary?
A general IT skills shortage no longer describes the Hungarian market as a whole. Employment and recruitment data point to softer demand than before, although some specialisms are still hard to hire for.
According to Eurostat, 212,500 information and communication technology (ICT) specialists worked in Hungary in 2025, 4.6% of total employment. That is far above the 151,000 of 2019, but growth versus 2024 was just 0.7%. (Eurostat)
The share of Hungarian companies recruiting ICT specialists fell from 8.25% in 2022 to 6.35% in 2024. Over the same period the share reporting hard-to-fill ICT vacancies dropped from 5.05% to 3.96%. Both figures were below the EU average. (Eurostat)
Share of Hungarian enterprises. Source: Eurostat isoc_ske_itrcrn2.
Official vacancy data point the same way. According to the Hungarian Central Statistical Office (KSH), the IT and telecoms sector had 1,916 unfilled vacancies in Q2 2026, 62% fewer than the 4,991 in Q2 2022. The vacancy rate fell from 5.0% to 1.8%, now below the national average of 2.1%. Most of that drop happened by the end of 2023, before AI coding tools were widespread. (KSH)
NACE Rev. 2.1 section K (telecoms, programming, IT consulting, infrastructure and other information services). Source: KSH table 20.2.1.55.
Across the Hungarian economy, unfilled vacancies fell by 31% between Q2 2022 and Q2 2026. The IT and telecoms sector fell twice as fast; only construction dropped more. In August 2026 IT was also one of the few sectors where headcount shrank year on year (−1.2%), and it was the best-paid sector in the country.
NACE Rev. 2.1 sections. Source: KSH table 20.2.1.55.
Full-time employees, million HUF (about €3,300 for IT at the current exchange rate). Source: KSH table 20.2.1.46.
Hungarian companies and AI: by sector, size and technology
According to Eurostat, 10.4% of Hungarian enterprises with 10 or more employees used AI in 2025, against an EU average of 20.0%, ranking Hungary 22nd of 27. The gap shows up in every sector and size class. IT service firms lead (51% vs 66% in the EU); in manufacturing only about one firm in thirteen uses AI. Among large companies the rate is 40%, among firms with 10–49 staff just 8.7%. (Eurostat)
Enterprises with 10+ employees. Source: Eurostat isoc_eb_ain2.
Persons employed. Source: Eurostat isoc_eb_ai.
Share of enterprises using each. Source: Eurostat isoc_eb_ai.
Among companies that had considered AI, the most common reason for not using it was a lack of expertise (65.5% in Hungary). One difference stands out: 29.8% of Hungarian firms said AI simply isn’t useful for them, against 17.8% across the EU.
Enterprises that had considered using AI; multiple answers allowed. Source: Eurostat isoc_eb_ai.
In June 2026 IVSZ, the Hungarian IT industry association, wrote that there are not enough digital specialists either in Hungary or in the EU, and that easing the ICT skills shortage “cannot be postponed any longer”. That shortage is measured against the 2030 digitalisation goals: the EU wants 20 million ICT specialists by 2030, up from about 10.5 million today. Current job postings show much softer demand than that target implies. (IVSZ, in Hungarian)
Many Hungarian developers work in development and shared-service centres of international companies. Hiring and headcount decisions there are often made abroad, so local opportunities also depend on the group as a whole.
In Dev Radar’s Hungarian listings that state a level, junior roles make up 17.2%, almost twice the share in the international data. That is a better ratio, but the far smaller number of Hungarian listings should be kept in mind. Remote roles can widen the options if the employer accepts candidates from Hungary. Local openings are listed under jobs in Hungary.
9. Three possible scenarios to 2030
Longer-term forecasts are uncertain. The BLS projects 10% employment growth for software developers between 2025 and 2035. The previous year’s projection, covering a different period, was 15.8%, so the two numbers are not strictly comparable. (BLS)
In the World Economic Forum’s 2025 report, software and applications developers ranked fourth among the fastest-growing occupations to 2030. In the same survey, 40% of employers planned to reduce staff where AI can automate tasks. (WEF)
Several paths are therefore possible. The scenarios below describe possible outcomes, not firm predictions.
More projects, changing developer work
Because development gets cheaper, more companies commission custom software or extend existing systems. The number of developers may grow, while system design, reviewing AI output, security and operations take up more of the job.
Getting in may stay harder even in this case. Companies would need to build training paths that let juniors gain the experience needed for independent work while using AI tools.
Fewer jobs, smaller teams
Automation removes more work than new projects create. Companies run smaller teams, and demand for mid-level roles may fall too. The pay gap between developers with different skills could widen.
Goldman Sachs estimates that AI could displace 6–7% of US workers over the longer term, with the transition raising unemployment by 0.5 percentage points. That estimate covers the whole labor market, not developers specifically.
Automation of larger development tasks
If models keep improving at the pace seen in METR’s measurements, well-specified development tasks lasting several days could become automatable. That could reshape parts of development well beyond writing code.
Projecting the current growth rate years ahead is uncertain, though. Solving benchmark tasks does not necessarily translate into the same reliability inside a company’s complex, constantly changing systems.
In all three cases it helps if a developer understands, beyond writing code, the business need, how the system works and the risks of a solution. That does not guarantee any particular job, but it builds knowledge that is useful across many roles.
10. What to focus on as a junior, mid-level or senior developer
If you are starting out or switching careers
- Learn to use AI tools, and check the code they write. In interviews you will need to explain how a solution works and where it could break.
- Build an independent project, ideally with real users, or contribute to an open-source project. You will run into decisions and bugs that tutorials rarely cover. Dev Radar’s GitHub profile preview shows which jobs your profile already matches.
- Look for internships and junior roles, and check what onboarding and mentoring the company offers. You can filter for junior and internship roles.
- Pick one direction in which you can build a working application on your own, for example Python or TypeScript, a framework, a database and the basics of deployment.
If you are mid-level
- Take on work that requires understanding a whole system: design, operations, security or performance problems.
- Find out where AI actually saves you time, counting review and fixes. Show your experience on your CV with concrete results.
- Compare your pay with the salary ranges of jobs asking for similar technologies and experience. Dev Radar’s salary calculator gives a baseline from advertised pay.
If you are senior
- Most of the growth in US postings comes from senior roles, but track demand in your own specialism and target countries separately.
- Beyond using AI tools, it can pay to know how to build LLMs into applications, RAG, AI agent design, and how to test and secure them. Higher advertised pay does not mean an automatic raise for each such skill. Roles asking for them: AI jobs.
- Help train juniors. If the team only ever hires finished professionals, building a pipeline later gets harder.
Dev Radar aggregates 69,115 active tech job listings directly from employers’ applicant tracking systems. Filter by technology, seniority, country, remote work and salary range, no registration needed. You apply on the employer’s own site.
Frequently asked questions
Will AI replace software developers?+
Research so far shows no broad, AI-driven unemployment among software developers. Entry-level developers are the exception: a Stanford analysis of payroll data (August 2026) finds that employment of 22–25-year-olds in the most AI-exposed occupations is 19% below the level implied by less-exposed groups, mostly because fewer new people are being hired. The US Bureau of Labor Statistics projects 10% growth for software developers from 2025 to 2035, but a 7% decline for the narrower “computer programmer” category.
Why is it so hard to get a junior developer job in 2026?+
Many tasks traditionally given to juniors (well-scoped tickets, boilerplate, simple bug fixes, tests) can now be partly handled with AI tools, so some teams absorb the work with existing staff and postpone new hires. A Harvard study of about 65 million résumés found junior employment at AI-adopting firms fell by roughly 9% over six quarters relative to non-adopters, with no comparable change for seniors. According to Indeed, 69% of software development postings in the first quarter of 2026 were for senior roles.
Does AI actually make developers faster?+
It depends on the task and the environment. In METR’s 2025 randomized trial, experienced open-source developers took 19% longer on their own codebases when using AI, even though they believed it had sped them up by 20%. A February 2026 follow-up measured shorter completion times, but METR rated that evidence as weak. Google’s DORA 2025 report links AI adoption to faster delivery but lower stability.
What skills should developers learn to stay employable?+
According to the World Economic Forum’s 2025 survey, the fastest-growing skills to 2030 are AI and big data, networks and cybersecurity, and technological literacy; employers expect 39% of core skills to change by 2030. In Stack Overflow’s 2025 survey the most-used languages were JavaScript (66%), HTML/CSS, SQL and Python (58%, up 7 points in a year).
Is there still a developer shortage in Hungary and the EU?+
Demand has cooled. According to Eurostat, 212,500 ICT specialists worked in Hungary in 2025 (4.6% of employment), but growth versus 2024 was only 0.7%. The share of Hungarian companies recruiting ICT specialists fell from 8.25% in 2022 to 6.35% in 2024. Official vacancy data from KSH show 1,916 unfilled vacancies in the IT and telecoms sector in Q2 2026, 62% fewer than in Q2 2022. IVSZ, the Hungarian IT industry association, still warns that there are not enough digital specialists for the 2030 digitalisation goals.
Where can I find software developer jobs?+
Dev Radar (devradar.hu) aggregates 69,115 active tech job listings directly from employers’ applicant tracking systems, 9,717 of them in AI or machine learning. You can filter by technology, seniority, country, remote work and salary range without registering, and you apply on the employer’s own site.
Methodology
This article draws on research, official statistics and company statements. External figures were checked against the primary source on October 19, 2026; where a study has appeared in several versions, we cite the version named and note the change. Numbers that companies report about themselves are labelled as such.
The small-print “own sample” notes come from Dev Radar’s job listings database, updated daily (state on November 22, 2026). It is not a representative sample: it mostly collects from international employers’ applicant tracking systems, so it is used only as a supplement.
Sources
- Brynjolfsson, Chandar, Chen: Canaries in the Coal Mine (Stanford Digital Economy Lab, August 2026 version)
- Hosseini, Lichtinger: Generative AI as Seniority-Biased Technological Change (SSRN, May 2026)
- Anthropic: Labor market impacts (March 2026) · Impact on software development (2025)
- Federal Reserve Bank of Dallas (January 2026)
- Indeed / FRED software development postings index · Indeed Hiring Lab (July 2026)
- METR randomized trial (2025) · METR follow-up (2026) · METR time horizon
- Google DORA 2025
- Stack Overflow Developer Survey 2025 · 2026
- BLS: Software Developers · Computer Programmers
- World Economic Forum: Future of Jobs Report 2025
- PwC Global AI Jobs Barometer 2026 (press release)
- layoffs.fyi · TechCrunch: 2026 layoffs citing AI · Amazon (Jassy, 2025)
- Google Cloud Next 2026 (Pichai)
- BLS JOLTS (information sector) · Eurostat jvs_q_nace2 · isoc_eb_ain2 · isoc_eb_ai
- Eurostat isoc_sks_itspt · isoc_ske_itrcrn2 · IVSZ (June 23, 2026) · KSH vacancies
- Goldman Sachs: How will AI affect the global workforce?
- Stack Overflow 2025: technology · work and AI
- Dev Radar (November 22, 2026): 69,115 active listings, 9,717 in AI; 36,994 listings state a seniority level.
Related: State of AI Dev Salaries 2026 · Read this article in Hungarian