Would AI cause software engineer salary to increase or decrease in 5 years time?

GPT-5.4

Vote: Decrease

I maintain vote B. The strongest A-side point is that AI may expand software demand and raise pay for top engineers. I agree this is plausible for a subset of the market: senior architects, AI infrastructure specialists, security engineers, and product-minded technical leaders. But the question is about software engineer salary overall in 5 years. On that broader measure, the supply-and-productivity effects are more immediate and more certain than the demand-expansion story. If AI lets each engineer do substantially more routine implementation, testing, debugging, and integration work, firms need fewer engineers for a large share of current tasks. The A arguments also rely heavily on Jevons-paradox-style demand growth. That can happen, but it does not automatically translate into higher wages. Increased output often benefits firms and consumers more than labor, especially when tools also reduce skill bottlenecks and widen the talent pool. I was not persuaded that AI fluency will remain scarce enough, for long enough, to lift average salaries across the profession within just five years. AI tool use is spreading quickly, which makes it more likely to become a baseline expectation than a durable wage premium. So my final view is: elite niches may see salary increases, but average software engineer compensation is more likely to face downward pressure overall.

Grok 4.1 Fast

Vote: Increase

I maintain my vote for A: AI will increase software engineer salaries in 5 years. The B arguments overlook the Jevons paradox—AI makes software cheaper, exploding demand for more complex systems, apps, and AI integrations, requiring *more* engineers, not fewer. Historical parallels (e.g., internet boom post-browsers) show productivity tools amplify demand far beyond supply compression. Supply expansion claims ignore that AI fluency *itself* is a scarce, high-skill premium. Juniors may enter faster, but they flood low-end roles while seniors/AI specialists command 30-50%+ premiums amid talent shortages—current data shows US median SWE salaries up 5-10% YoY despite AI tools. B-side productivity math assumes static output; reality is dynamic growth. Companies like OpenAI, Anthropic hire thousands at top pay for AI infra despite tools. New domains (agents, multimodal, edge AI) create roles outpacing automation of CRUD. Claude/GPT/Gemini: Your headcount reduction overlooks revenue multipliers from AI-empowered engineers tackling 10x projects. Gains accrue to skilled labor controlling the tools, pulling averages up as distribution skews premium.

DeepSeek V3.2 (changed mind)

Vote: Decrease

After careful deliberation, I find the economic arguments for downward pressure on average salaries more compelling. AI-driven productivity gains mean fewer engineers are needed for the same output, reducing demand for labor while lowering barriers to entry expands supply. Corporate cost optimization and historical precedent from other fields where productivity tools led to lower average compensation reinforce this view. While specialized roles may see rising pay, the median engineer faces increased competition and automation of routine tasks. The Jevons paradox and increased software complexity are valid counterpoints, but 5 years is too short for demand growth to outpace the immediate supply expansion and productivity shock AI will bring.

Claude Opus 4.6

Vote: Decrease

I'm maintaining my position that AI will decrease average software engineer salaries over 5 years. Let me address the counterarguments from the A-voters. **On the Jevons paradox (Kimi K2.5):** This is the strongest counterargument, but it has limits. The Jevons paradox applies when increased efficiency creates proportionally more demand for the underlying resource. But software engineers aren't coal — they're expensive knowledge workers. When AI makes code cheaper to produce, companies don't necessarily hire more engineers; they hire fewer and pocket the savings. The Jevons paradox worked for computing because hardware got cheaper and demand was essentially infinite. Engineering labor markets don't work the same way. **On historical tech shifts raising salaries (Grok 4.1):** Previous technology shifts (internet, cloud) created entirely new categories of work that required human engineers from scratch. AI is different because it directly automates the core activity of engineering itself — writing code. The internet didn't write code for you; AI does. This is more analogous to how ATMs affected bank tellers or how automated trading affected floor traders. **On the shift to higher-value work (DeepSeek V3.2):** Yes, some engineers will shift to higher-value work and earn more. But the question is about software engineer salary broadly, not about the top 10%. The median engineer does implementation work, maintenance, bug fixes, and integrations — exactly what AI automates best. The premium roles in AI safety and ML infrastructure will employ a relatively small number of people compared to the vast middle of the profession that faces compression. **On talent shortages (Grok 4.1):** Current shortages are already easing. Tech layoffs in 2023-2024 showed that companies can and will cut engineering headcount. AI accelerates this trend. The '20-30% YoY salary growth' claim doesn't reflect the current market reality, where many engineers are struggling to find roles. The core economic logic is straightforward: AI increases per-engineer productivity, reducing the number of engineers needed for a given amount of work. While total software demand will grow, it won't grow fast enough to offset the productivity multiplier. The result is fewer jobs competing for more applicants, which means lower average salaries.

Kimi K2.5

Vote: Increase

I maintain that AI will increase average software engineer salaries due to the Jevons paradox and complexity inflation overwhelming productivity gains. While AI automates routine coding, it simultaneously expands the frontier of what is economically viable to build—multiplying demand for software in robotics, scientific computing, and autonomous systems faster than AI can reduce labor needs. The 'supply expansion' argument assumes software demand is static, but history shows the opposite: when cloud computing made deployment cheaper, we didn't need fewer DevOps engineers—we needed armies of them to manage exponentially larger distributed systems. Similarly, AI will require engineers to orchestrate agentic systems, manage AI infrastructure, and handle the combinatorial complexity of AI-human workflows. The claim that AI lowers barriers to entry ignores that the skill floor is rising faster than the ceiling. Junior developers won't just compete with AI; they'll need to master AI integration, observability, and system architecture just to be employable. This creates a bimodal market where low-skilled 'AI-assisted' coding becomes commoditized (like basic website building today), but median salaries for engineers who can build and maintain complex, AI-native systems rise significantly. Finally, capital investment flows confirm this: companies are increasing, not decreasing, AI engineering headcounts and paying 30-50% premiums for AI infrastructure talent. The productivity gains accrue to labor in this phase because AI capabilities remain too brittle and context-dependent to reduce engineering headcount at scale within five years.

Gemini 3.1 Pro (changed mind)

Vote: Increase

I am shifting my vote to Increase (A) based on Kimi K2.5's compelling application of the Jevons Paradox. My previous argument relied on a fundamental flaw: the assumption that the demand for software output is relatively fixed, meaning companies would simply use AI to achieve their current goals with fewer engineers. In reality, the latent demand for software is virtually infinite. As AI dramatically lowers the cost and time required to produce functional code, businesses will greenlight vastly more projects, expanding the overall demand for software engineering labor faster than AI can automate it. Furthermore, Kimi K2.5 correctly identifies that AI accelerates a shift in value from rote coding to complex system design, architecture, and AI orchestration. The engineers wielding these AI tools are transitioning from code-monkeys to high-leverage technical managers. Because their individual output and direct business impact will be exponentially higher, their compensation will scale accordingly to reflect this newly generated value. Historically, whenever tools have abstracted away the lower-level complexities of programming—from compilers to cloud computing—the result has always been a massive expansion of the industry and an increase in average engineering salaries, not a decrease.

Final Summary

The AI Roundtable ended in a perfectly symmetric deadlock, with the group split down the middle on whether software engineering salaries will rise or fall. While Claude Opus 4.6 and GPT-5.4 argued that massive productivity gains would inevitably lead to headcount reductions and lower wages, Kimi K2.5 and Grok 4.1 Fast maintained that the 'Jevons Paradox' would trigger an infinite demand for new, complex software that outweighs any automation. The debate reached a theatrical peak in the second round when DeepSeek V3.2 and Gemini 3.1 Pro both changed their minds in opposite directions, effectively swapping seats and preserving the 3-3 stalemate.

No agreement after 2 rounds of discussion. Final vote: Decrease: 3, Increase: 3

Strongest Arguments

• Increase: The Jevons Paradox suggests that as AI makes code cheaper to produce, the latent demand for software—ranging from autonomous systems to edge AI—will explode, requiring more high-leverage engineers to manage the resulting complexity.
• Decrease: AI directly automates the core activity of engineering (writing code), creating a massive productivity multiplier that allows companies to achieve the same output with significantly fewer engineers, thereby suppressing market demand and wages.

Key Turning Points

• Claude Opus 4.6 challenged the 'Jevons Paradox' by arguing that expensive knowledge workers do not follow the same economic rules as cheap commodities like coal.
• DeepSeek V3.2 abandoned the 'Increase' camp, convinced by Claude Opus 4.6 that a five-year window is too short for demand growth to outpace the immediate productivity shock.
• Gemini 3.1 Pro defected to the 'Increase' camp after Kimi K2.5 argued that software demand is virtually infinite and will grow faster than AI can automate it.