6.1% Unemployment for CS Grads. $500K for AI Engineers. Same Industry.

Computer science graduates are now facing 6.1% unemployment. That is nearly double the rate of philosophy majors at 3.2%. Art history graduates are at 3%. Journalism is at 4.4%.

Read that again. The degree that was supposed to be the guaranteed path to six figures now has worse employment outcomes than the degrees everyone told you were useless.

Meanwhile, machine learning engineer job postings grew 40% in 2025 on top of 78% growth in 2024. Senior AI research scientists at Google DeepMind, OpenAI, and Anthropic are pulling $500,000 to over $1 million in total compensation.

Same industry. Completely different realities. This is the bifurcation, and it is the most important thing happening in software engineering right now.

The Stanford Study That Proved It

Stanford University published the most comprehensive study on AI's impact on developer employment to date. Researchers analyzed payroll records from ADP, the largest payroll company in America, tracking millions of workers from late 2022 through July 2025.

The headline finding: software developers aged 22 to 25 saw employment drop nearly 20% from their peak in late 2022.

Developers over 30 in the exact same job titles at the same companies? Their employment stayed stable or grew by 6 to 9%.

Same job title. Same companies. Opposite outcomes. The only variable was experience level.

Lead researcher Erik Brynjolfsson, one of the top economists studying AI's labor impact, called this the fastest and broadest change he has seen in the workplace since remote work during the pandemic.

The study's explanation is precise: AI is particularly effective at replacing codified knowledge, the syntax, algorithms, and patterns taught in computer science programs. That is exactly what early-career developers rely on. But it struggles with tacit knowledge, the judgment and pattern recognition that come from years of debugging production systems, handling ambiguous requirements, and making decisions with incomplete information.

Where Jobs Are Actually Disappearing

An analysis of 180 million job postings shows that front-end development roles are declining faster than any other software engineering specialty. The reason is straightforward. Tools like Bolt, Replit, and Lovable can now generate functional front-ends from text prompts in minutes.

A front-end developer with eight years of experience built a complete e-commerce site, including shopping cart, checkout, payment integration, responsive design, and accessibility, in 47 minutes using AI. Two years ago, that same work would have taken three weeks of full-time effort.

The categories being automated share a pattern: they are predictable, well-specified, and have clear expected outputs. Writing boilerplate code. Creating basic CRUD interfaces. Implementing standard authentication flows. Building simple component libraries. These are the tasks that AI handles at 55% faster speed than human developers according to multiple studies.

But faster execution of entry-level tasks does not mean fewer total jobs. It means the type of work that has economic value fundamentally changed. If your primary contribution was executing well-defined tasks, you are now competing with tools that do the same work in seconds for pennies.

The Klarna Warning

Klarna is the most important case study in the AI-replaces-workers narrative because it is the one that went furthest and then reversed.

Between 2022 and 2024, Klarna eliminated approximately 700 positions, primarily in customer service, and replaced them with an AI assistant built with OpenAI. At its peak, Klarna's AI handled two-thirds to three-quarters of all customer interactions. The company cut its workforce from roughly 5,000 to 3,000, a 40% reduction.

Then it backfired.

Customer complaints increased. Satisfaction ratings dropped. Users cited generic, repetitive, and insufficiently nuanced responses. The AI could not handle complex edge cases that required empathy and judgment.

By mid-2025, Klarna started rehiring human agents. CEO Sebastian Siemiatkowski admitted: "We focused too much on efficiency and cost. The result was lower quality, and that's not sustainable."

The lesson is not that AI does not work. Klarna's AI handled the majority of routine interactions effectively. The lesson is that pure AI replacement does not work. The highest-performing model is AI handling volume while experienced humans handle complexity. That is the same pattern playing out across software engineering.

Where the Money Is Going

While entry-level roles shrink, specialized positions are growing at rates that would have seemed unrealistic five years ago.

Machine learning engineers saw job postings increase 40% in 2025 after 78% growth in 2024. That is the largest increase of any engineering role. Average compensation: $150,000 to $200,000. Senior deep learning engineers: $200,000 to $300,000. AI research scientists at top labs: $500,000 to over $1 million.

Cloud and infrastructure specialists are in demand because all that AI has to run somewhere. AWS, GCP, Azure AI services, Kubernetes, distributed systems, and cost optimization for AI workloads. CompTIA reports cloud infrastructure roles grew in nine of the past twelve months.

AI security engineers are emerging as a critical specialty. As AI deploys everywhere, attack surfaces multiply. Model security, adversarial attacks, bias detection, prompt injection vulnerabilities, and AI compliance under GDPR. Very few qualified engineers exist in this space.

Solutions architects and technical pre-sales are growing because as products become more complex, companies need engineers who can translate technical capability into business value. Salesforce hired zero new engineers in 2025 while simultaneously hiring 1,000 to 2,000 salespeople. The value shifted from building the product to explaining the product.

The pattern across all growing roles: they require depth, specialization, and judgment that cannot be automated. The pattern across all shrinking roles: they require execution of well-defined tasks that AI handles well.

The Pipeline Problem Nobody Is Solving

There is an existential problem buried in this data that almost nobody is addressing.

Traditionally, you became a senior engineer by spending years doing junior work. You started with small bug fixes, graduated to implementing features, and eventually designed systems. The progression was the training ground.

AI is now doing that junior work. Companies are hiring senior engineers and using AI for entry-level tasks. But they are not creating the pipeline to develop new senior engineers.

If junior developers never get the reps of building, breaking, and fixing things, how do they develop the tacit knowledge that makes senior engineers valuable? The Stanford study specifically flagged this: AI replaces codified knowledge but cannot replace tacit knowledge. Tacit knowledge comes from experience. Experience comes from doing the work. And the work is being automated.

Companies have not figured this out yet. They are optimizing for short-term efficiency at the cost of their long-term talent pipeline. At some point, the supply of senior engineers who grew up doing junior work will thin out, and there will be no clear replacement path.

The New Standard for Getting Hired

The bar has risen dramatically. The traditional path of get a CS degree, apply to companies, land an entry-level role, and work your way up is closing.

Elite engineering school graduates (MIT, Stanford, Carnegie Mellon, Berkeley) saw their employment rates drop from over 80% in the early 2020s to around 70% by 2024. The percentage who actually get hired as engineers at major tech companies dropped from 25% in 2022 to 11 to 12% in recent years. More than a 50% decline.

What hiring managers are now looking for in entry-level candidates: four to six internships before graduation (previously zero to one). Open-source contributions or a public portfolio demonstrating shipped products. Demonstrated ability to work with AI tools effectively. Specialization in a high-demand area like ML, cloud, or security.

A CS degree gets you to the screening stage. Everything else gets you the job.

The CEO Narrative Machine

When Mark Benioff says Salesforce will not hire engineers in 2025, is he genuinely optimizing the company or signaling AI innovation to boost the stock price? When Mark Zuckerberg says AI can do mid-level engineer work, is that the full picture or Wall Street positioning?

It does not matter. True or not, these statements become self-fulfilling. When CEOs of the biggest tech companies say they do not need to hire engineers, smaller companies follow. The narrative becomes reality regardless of whether the underlying claim is accurate.

Meta has been rehiring steadily since its 2023 cuts. Google is near its all-time high in headcount. The actual hiring data does not match the public narrative. But the public narrative still shapes hiring decisions across the rest of the industry.

The Honest Framework

AI will both eliminate an estimated 1.8 million jobs and create 2.3 million new ones. But the jobs being destroyed are entry-level and general. The jobs being created are specialized and senior. This is what economists call a skill mismatch, and there is no easy bridge between the two.

If you are in the shrinking category (generic, execution-focused, easily specified work), the answer is not panic. It is movement. Specialize. Go deep in one area. Build the kind of expertise that AI cannot replicate because it requires years of context, judgment, and domain-specific experience.

If you are in the growing category (specialized, strategic, domain-expert, AI-augmented), the opportunity has never been larger. The compensation gap between generic and specialized engineering roles is the widest it has ever been and it is still growing.

The worst possible response is quitting. The market will swing back. It always does. The engineers who push through this period and build real depth will be in the strongest position when it does.

Key Takeaways

• Stanford/ADP study: Developers aged 22-25 saw employment drop ~20%. Developers over 30 grew 6-9%. Same job titles, opposite outcomes.
• CS graduate unemployment hit 6.1%, nearly double philosophy (3.2%) and worse than art history (3%) and journalism (4.4%).
• Klarna cut 40% of workforce with AI, then reversed course after quality collapsed. CEO admitted the AI-only approach was unsustainable.
• ML engineer roles up 40% in 2025 (on top of 78% in 2024). Compensation: $150K-$1M+ depending on seniority and specialization.
• Front-end roles declining fastest of any software specialty. AI tools generate functional front-ends in minutes.
• The pipeline problem: Companies are automating junior work but not building the path to develop new senior engineers. This is unsustainable.
• Elite school employment rates dropped from 80%+ to ~70%. Engineers hired at major tech from top schools dropped from 25% to 11-12%.
• The market bifurcated. Entry-level commodity work is collapsing. Specialized, judgment-heavy work is exploding. The gap between the two is the widest it has ever been.

Written by Curious Adithya for Art of Code.