OpinionMarch 6, 2026

Against the Tide: AI Isn't Killing Software Jobs — The Data Tells a Different Story

FRED data shows software dev job postings have stabilized since mid-2024, challenging the narrative of AI-driven collapse.

The headlines are unrelenting: AI is coming for software developers. ChatGPT writes code now. GitHub Copilot is replacing junior engineers. The profession, we're told, is on borrowed time.

But what does the data actually say?

The FRED Index: A Reality Check

The U.S. Bureau of Labor Statistics, via the FRED (Federal Reserve Economic Data) database, tracks job postings for software development occupations through the Indeed Job Postings Index (series IHLIDXUSTPSOFTDEVE). As of early 2025, the index sits at approximately 69.88, with February 2020 (pre-pandemic) set as the baseline of 100.

Yes, that's a ~30% decline from pre-pandemic levels. But here's what the doom narratives conveniently omit: the index has been largely flat since mid-2024.

The Stabilization Nobody's Talking About

After a sharp correction from the pandemic-era hiring frenzy (when the index soared well above 100), software development job postings have found a floor. The decline wasn't a cliff — it was a normalization. Consider the context:

  • 2021-2022: Tech companies hired at unsustainable rates, fueled by zero-interest-rate stimulus and pandemic-driven digital acceleration
  • 2023: The correction hit — mass layoffs at Meta, Google, Amazon, and Microsoft as companies right-sized after over-hiring
  • 2024-2025: The index stabilized around 65-70, suggesting the market has found its new equilibrium

This pattern looks less like "AI is destroying software jobs" and more like "the tech industry over-hired during a bubble and corrected."

What AI Is Actually Doing to Software Work

The nuanced truth is that AI is reshaping software development, not eliminating it:

  • Productivity gains: Developers using AI tools report 25-55% productivity improvements (GitHub's own research). This means the same output requires fewer hours — but it also means individual developers can tackle more ambitious projects
  • Shifting skill demands: The fastest-growing software roles in 2025 are in AI/ML engineering, platform engineering, and AI safety — all require deep software expertise
  • The "10x developer" effect: AI tools are amplifying the gap between strong and weak developers, changing who gets hired rather than whether people get hired

The Denominator Problem

Critics citing the FRED decline forget a crucial denominator: the total number of software developers has grown consistently. The BLS projects software developer employment to grow 25% from 2022 to 2032 — far faster than average. Fewer job postings doesn't mean fewer jobs. It can also mean:

  • Lower turnover (people staying in roles longer)
  • More internal mobility (companies retraining existing staff)
  • Faster hiring cycles (positions filled more quickly with better matching)

The Real Risk Isn't Replacement — It's Polarization

The legitimate concern isn't that AI eliminates all software jobs. It's that AI creates a two-tier labor market:

  1. Top tier: Senior developers, AI specialists, and architects who leverage AI tools to deliver 10x value — commanding premium salaries
  2. Compressed tier: Junior and mid-level developers in routine roles facing genuine wage pressure and competition

This is a real problem worth addressing through education and policy. But it's a fundamentally different narrative than "AI is killing software jobs."

What to Watch

Three indicators will tell us whether the stabilization holds or if a genuine decline materializes:

  • The FRED index trajectory through 2025: If it holds above 60, the stabilization thesis is confirmed
  • Computer science enrollment data: Currently still near record highs — students aren't buying the doom narrative
  • AI tool adoption curves: As AI assistants mature, watch whether they create net new demand for developers who can orchestrate these systems

The Bottom Line

The data suggests we're witnessing a market correction and transformation, not a collapse. The ~30% decline from pandemic highs is significant — but the stabilization at current levels tells a story of adaptation, not annihilation.

Before declaring the software profession dead, perhaps we should let the data finish its sentence.


Data source: FRED IHLIDXUSTPSOFTDEVE — Indeed Job Postings Index for Software Development Occupations