
Will AI Replace Software engineers?
AI already writes a large share of new code at several big tech firms, and some have cut staff or frozen engineering hiring. Overall demand for developers is still projected to grow, but early-career engineers are taking the hit first.
The fear is simple: AI now writes so much code that companies will stop paying people to write it. The evidence points somewhere narrower. AI is cutting junior hiring and team sizes at some firms, while total demand for developers is still projected to rise.
Start with the official baseline. The BLS employment projections for software developers count 1,717,800 jobs in 2025 and project 1,892,600 for 2035. That is a +10.2% change, against 3.5% for all occupations, with 95,300 openings a year on average and a $135,980 median annual wage in 2025. A bachelor's degree is the typical entry credential.
Earlier BLS work pointed the same way. A BLS Monthly Labor Review case study on AI impacts projected 17.9 percent growth for software developers between 2023 and 2033 and noted that "AI itself may lead to increased demand for software developers". The BLS 2024 to 2034 outlook projected growth of 15.8 percent, an increase of over 267,000 jobs. The projected growth rate has come down with each round, but it has stayed well above the all-occupation average.
Myths vs reality: what the AI coding claims actually show
| Myth | Reality |
|---|---|
| AI writes most of the code, so engineers are finished. | Snap says AI agents generate over 65% of its new code, per CNBC. Yet BLS still projects a +10.2% rise in developer jobs from 2025 to 2035. |
| AI makes every developer faster. | In a METR randomized trial, experienced open-source developers took 19% longer with AI tools. METR now labels that result as historical, according to ScienceBlog, and earlier research cited by Yahoo Tech found a 56% speedup. Results depend heavily on the task and the codebase. |
| Code that runs is code you can ship. | Veracode tested 80 curated coding tasks across more than 100 large language models and found functional code that still carried security flaws. Its March 2026 follow-up, summarized by the Cloud Security Alliance, found larger models did not outperform smaller ones on security. |
| Every tech layoff is an AI layoff. | Analyst rollups cited by CNBC attribute about 25% of March 2026 cuts to AI and automation and roughly 75% to cost discipline and restructuring, per SQ Magazine. |
Where AI is already writing code and shrinking engineering teams
This is not a forecast. Several large employers have publicly tied code output, hiring or layoffs to AI tools.
- Snap, April 2026. Snap is cutting roughly 16% of its global workforce, around 1,000 full-time employees, and closing more than 300 open roles, TechCrunch reports. CEO Evan Spiegel's memo credits "rapid advancements in artificial intelligence" for the change. CNBC adds that the cuts target more than $500 million in annualized cost savings and that the company has seen "small squads leveraging AI tools to drive meaningful progress".
- Salesforce, fiscal year 2026. The company hired zero new engineers, with Marc Benioff citing AI coding tools, according to SQ Magazine. We could not check this against a primary source.
- Microsoft. Satya Nadella's claim of 20% to 30% AI-generated code applies to select coding projects, not the full codebase, the same outlet notes.
- Google, Q4 2025 earnings call. Sundar Pichai said AI generates over 30% of all new code at the company, per Sundeep Teki. Another aggregator reports a much higher Google figure, so treat any single number with caution.
- Airbnb and Uber, 2026. Airbnb's CEO said on a Q1 earnings call that AI writes 60% of the company's new code, and Uber engineers reached 70% AI involvement in committed code by April 2026, Grey Journal reports.
- Shopify, April 2025. CEO Tobi Lutke's memo made AI usage a fundamental expectation for staff, per Grey Journal.
- Meta, 2026. Reported leaked memos set targets for AI-written code inside Meta, and Meta cut about 10% of staff in April 2026 in what SQ Magazine labels AI restructuring. Meta pushes back: "The company says these layoffs and the AI mandates are unrelated," HappyCapy Guide notes.
The tools are also getting more independent. Anthropic says that "Claude models can operate autonomously for hours through Claude Code and Cowork," in its June 2026 Economic Index report. Its software development study found 79% of Claude Code conversations involved some form of automation, compared to 49% on Claude.ai.
Most of these share-of-code numbers come from aggregators and executive remarks, not audited data. They measure lines generated, not engineers made redundant.
Skills that gain value as AI writes more of the code
When a machine drafts the code, the scarce work becomes judging it. The evidence points to a few specific areas.
- Secure code review. Veracode's findings mean someone has to catch the flaws AI introduces. Bigger models did not fix this, so security review skill does not lose value as tools improve.
- Owning what ships. The Linux kernel's guidelines on tool-generated content tell contributors to "expect additional scrutiny in proportion to how much of it was generated." PC Guide reports that "AI agents MUST NOT add Signed-off-by tags." A human signs and answers for the code.
- Building AI systems themselves. The BLS case study expects developers to be needed to build and maintain AI-based business solutions.
- Licensing and provenance awareness. On September 16, 2026 the Ninth Circuit affirmed dismissal of a DMCA claim in Doe v. GitHub, Haynes Boone reports. But "The Ninth Circuit did not decide whether copyrighted code can legally be used to train AI systems," Lexology notes. Engineers who understand open-source license terms stay useful while that question remains open.
What AI does in a software engineer's week, task by task
| Task | What AI does today | What still needs a person |
|---|---|---|
| Writing new feature code | Drafts large shares of new code at firms like Snap, Airbnb and Uber. | Deciding what to build, and rejecting drafts that do not fit the system. |
| Repetitive or boilerplate work | Snap says AI lets teams reduce repetitive work. | Spotting when repetition hides a design problem worth fixing. |
| Security review | Produces functional code that can still fail security checks, per Veracode. | Threat modeling and catching vulnerabilities before release. |
| Changes in large, mature codebases | Gave experienced developers mixed results; METR measured a slowdown in early 2025. | Deep context on how a long-lived system actually behaves. |
| Multi-step agent tasks | Coding agents can run autonomously for hours, per Anthropic. | Setting scope, checking output and stopping runs that drift. |
| Submitting and signing off code | Can generate contributions, flagged with an Assisted-by tag in the kernel. | Legal and technical responsibility for every merged change. |
Who is most exposed inside software engineering
New graduates face the sharpest drop. The number of software developers aged 22 to 25 fell by nearly 20 percent as of July 2025 compared to its peak in late 2022, according to a Stanford study reported by Breitbart. CNBC lists software development among the most AI-exposed jobs for this age group.
Experienced engineers are holding steady so far. Employment for more experienced workers in the same jobs has been stable, The Register reports.
Pure coding roles are more exposed than design roles. BLS projects a 7% fall in employment for computer programmers from 2025 to 2035 as companies automate repetitive programming, per Second Talent. Software developers, who also design and plan systems, get the growth forecast instead.
Startup engineers will feel it first. "Startups are the main early adopters of Claude Code, while enterprises lag behind," Anthropic found. Large enterprise teams may have more time to adapt.
Ratings differ on overall exposure. Extern reports that BLS rated software developers very high on an AI exposure scale published in August 2026, while AI Job Impact Calculator cites an ILO 2025 ranking that places them in a moderate band. Coding is also the top use of Anthropic's tools, at 35% of conversations, its Economic Index report shows. High exposure means the work changes. It does not by itself mean the job vanishes.
Our verdict: fewer junior seats, steady demand, more accountability
Our read: AI will not replace software engineers as a profession in the next decade, but it is already replacing some of the hiring that used to happen at the bottom of the ladder. The growth projections, the security record of AI code and the kernel's sign-off rules all point to engineers becoming reviewers and owners of machine output.
We think the people at real risk are those who only translate tickets into code. The people who gain are those who can judge, secure and take responsibility for what the tools produce.
Questions people ask
Will AI replace software engineers?
BLS projects software developer jobs to grow from 1,717,800 in 2025 to 1,892,600 in 2035, a +10.2% change. Our read is that AI reshapes the job and cuts junior hiring rather than eliminating the profession.
How much code does AI already write at big companies?
Snap says AI agents generate over 65% of its new code, Airbnb's CEO cited 60%, and Google's CEO said over 30% at the Q4 2025 earnings call. These figures come from executive remarks and aggregators and are not audited.
Are junior developers hit harder than senior ones?
Yes, so far. A Stanford study found software developers aged 22 to 25 dropped by nearly 20 percent as of July 2025 compared to late 2022, while employment for more experienced workers in the same jobs held stable.
Does AI make developers more productive?
It depends. A METR trial found experienced open-source developers took 19% longer with AI tools, though METR now calls that result historical, and an earlier study found a 56% speedup.
Is AI-generated code safe to ship?
Not without review. Veracode tested 80 coding tasks across more than 100 models and found working code that still had security flaws, and larger models did not perform better on security.
Sources
Every number, quote and example on this page was checked by script against the sources below before publishing: quotes must match the source word for word, and figures must appear in a source or in BLS data. Our verdict is editorial judgment. Spotted an error? Tell us.
- U.S. Bureau of Labor Statistics: Employment Projections, Software developers (15-1252)
- Software Engineer Layoff Statistics 2026: Companies, Roles, AI Impact (sqmagazine.co.uk)
- Impact of AI on the Software Engineering Job Market in 2026 (sundeepteki.org)
- How Many Engineers Does AI Replace in 2026 (greyjournal.net)
- Meta Mandates 75% AI-Written Code — Engineers Laid Off as Zuckerberg Says AI Replaces Whole Teams (happycapyguide.com)
- AI adoption linked to 13% decline in jobs for young U.S. workers, Stanford study reveals (cnbc.com)
- Stanford Study: AI Is Disrupting Entry-Level Jobs in Software Development and Other Fields (breitbart.com)
- AI robs jobs from recent college grads, no wage compression (theregister.com)
- Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity (metr.org)
- Experienced developers were 19 percent slower with AI tools in a METR trial (scienceblog.com)
- AI slows down some experienced software developers, study finds (tech.yahoo.com)
- Incorporating AI impacts in BLS employment projections: occupational case studies : Monthly Labor Review : U.S. Bureau of Labor Statistics (blsmon1.bls.gov)
- Will AI Replace Software Engineers? What the 2026 Data Shows | Extern (extern.com)
- Artificial intelligence, information technology, and employment, 2024–34 : The Economics Daily : U.S. Bureau of Labor Statistics (bls.gov)
- AI in Software Development Statistics for 2026 - Second Talent (secondtalent.com)
- Will AI Replace Software Developers? Risk Score, Tasks, Growth (2026) (aijobimpactcalculator.com)
- Snap's stock jumps on plans to axe 16% of its workforce citing AI efficiencies (cnbc.com)
- Snap is cutting 1,000 jobs, 16% of its workforce | TechCrunch (techcrunch.com)
- Anthropic Economic Index Report The Anthropic Economic Index report: (cdn.sanity.io)
- Anthropic Economic Index report: Cadences (anthropic.com)
- Economic Index: AI's impact on software development (anthropic.com)
- Ninth Circuit Limits DMCA Claim Against GitHub Copilot, But Leaves AI Training Question Open - Lexology (lexology.com)
- Ninth Circuit Rejects DMCA Copilot Theory | News | Haynes Boone (haynesboone.com)
- Vibe Coding’s Security Debt: The AI-Generated CVE Surge (labs.cloudsecurityalliance.org)
- AI-Generated Code Poses Major Security Risks in Nearly Half of All Development Tasks, Veracode Research Reveals (businesswire.com)
- Kernel Guidelines for Tool-Generated Content — The Linux Kernel documentation (cdn.kernel.org)
- Linux kernel now allows AI-generated code, as long as you take "full responsibility" for any bugs (pcguide.com)