Microsoft Cut 6,000 Jobs. Then It Kept Hiring AI Engineers.

In May 2025, Microsoft announced it was laying off approximately 6,000 employees — about 3% of its global workforce. The same quarter, Microsoft's job postings for AI engineers, Azure AI infrastructure roles, and GitHub Copilot product teams were accelerating. The two facts are not contradictory. They describe the same strategic decision: compress the workforce in roles AI has made more productive, reinvest in the roles building AI.

Microsoft is not unique. The same pattern ran through Google, Meta, and Amazon across 2023–2026. Understanding the actual numbers — who got cut, who got hired, what the net change was — tells you more about where developer careers are going than any general prediction about AI and employment.

Microsoft: The 6,000 Number in Context

Microsoft's May 2025 layoffs were concentrated in:

• Middle management layers, particularly in the non-AI product divisions
• Legacy Bing engineering teams consolidated into the AI-powered search organization
• LinkedIn roles in recruiting and content operations
• Gaming division roles following the Activision integration

Simultaneously, Microsoft was hiring aggressively in:

• Azure AI infrastructure (expanding capacity for OpenAI partnership workloads)
• GitHub Copilot and GitHub Models engineering
• Microsoft Research AI teams (safety, interpretability, new capability development)
• Enterprise AI sales engineering — the customer-facing technical roles that help companies deploy Copilot in enterprise environments

The net change in total Microsoft headcount from Q1 2025 to Q1 2026 was approximately flat to slight growth. The company did not become smaller. The composition shifted: fewer engineers maintaining legacy products, more engineers building AI-native products and infrastructure.

This matters because the narrative of "Microsoft laid off 6,000 people" and the narrative of "Microsoft is one of the largest hirers of AI talent in the world" are both true simultaneously and describe a workforce recomposition, not a workforce reduction.

Meta: The Same Math, Larger Scale

Meta's 2022–2023 layoffs were the largest in the company's history: approximately 21,000 employees cut across two rounds. Mark Zuckerberg explicitly described this as eliminating roles that AI would handle, funded by aggressive AI investment.

By 2025, Meta had rebuilt its headcount close to pre-layoff levels — but with a significantly different skill composition. The rebuild was concentrated in:

• AI research (FAIR — Fundamental AI Research — expanded substantially)
• AI infrastructure engineering (training clusters for Llama models)
• AI product engineering (Meta AI assistant, recommendation systems)
• Reality Labs hardware engineering

What did not rebuild: middle management (Meta ran "flatter" post-layoffs), large content moderation teams (AI handles more of this), and recruiting/HR functions scaled to the old headcount.

The productivity per employee metric improved significantly. Meta reported revenue per employee increasing from approximately $400,000 in 2022 to over $750,000 by 2024. Some of that is revenue growth; a meaningful portion is fewer people doing equivalent or more output because AI tools increased individual productivity.

Google: The Harder Case

Google's layoffs in January 2023 (12,000 employees, approximately 6% of workforce) were followed by a more complex hiring pattern than Microsoft or Meta. Google had historically operated with generous staffing and slow processes. The layoffs were partly efficiency-driven and partly a response to investor pressure after years of headcount growth without proportional revenue growth.

What makes Google's case harder to analyze: Google has both the products that AI threatens (Search, where AI-generated answers reduce ad impressions) and the infrastructure that AI requires (TPUs, data centers, Google Cloud). The divisions pointing in different directions are inside the same company.

Google Cloud — which benefits from AI demand — added engineering headcount throughout the layoff period. Google Search — which faces structural demand challenges from AI answer generation — contracted. The net company headcount was roughly flat but the internal redistribution was significant.

The developer-relevant takeaway: Google Cloud engineering roles have been consistently more insulated from layoffs than Google consumer product roles. If you are a developer with infrastructure, distributed systems, or ML infrastructure skills, Google Cloud is a different hiring environment than Google Maps or Google Assistant.

What the Pattern Means for Developers

The common thread across Microsoft, Meta, and Google: AI productivity gains are being taken partly as cost savings and partly as capability expansion. Companies are not choosing one or the other. They are doing both simultaneously, which is why headline layoff numbers coexist with aggressive AI hiring.

For developers, the specific role composition changes matter more than the headline numbers:

What is contracting:

• Roles where AI coding tools have dramatically increased individual productivity without requiring AI expertise — basic web development, internal tooling, standard CRUD applications
• Middle management in product organizations that had grown during the zero-interest-rate era
• Roles adjacent to products that AI is making less relevant (certain search team roles, basic content moderation)

What is expanding:

• AI infrastructure engineering: training clusters, inference optimization, GPU cluster management
• AI product engineering: building products that use AI as a core component rather than as an add-on feature
• AI safety and alignment engineering: regulatory pressure and company risk management are creating dedicated teams
• Enterprise AI deployment: sales engineers, solutions architects, and technical account managers helping large companies deploy AI tools

What is stable but recomposing:

• Senior software engineering and architecture: these roles have not declined significantly, but the expected AI tool fluency is higher
• DevOps and platform engineering: AI tools generate more code that needs deployment infrastructure, which sustains demand

The Salary Signal

The most reliable indicator of where the job market is repricing is the compensation differential between role categories. When a market values a skill, compensation premiums appear before headcount changes.

Current signals from public compensation data and job posting analysis:

• "AI engineer" roles at large tech companies average 20–35% higher base salary than "software engineer" roles at equivalent seniority levels
• Roles requiring ML infrastructure skills (GPU cluster management, distributed training, inference optimization) command premiums of 30–50% over equivalent DevOps roles
• Enterprise AI deployment roles (solutions engineers who understand AI products) are being filled at senior software engineer compensation, which is elevated relative to the historical positioning of solutions engineering

Roles seeing compression:

• Junior software engineering, where AI tools have reduced the supply-demand imbalance that drove junior salaries up during 2020–2022
• Content and social team engineering roles at large consumer platforms

The Long View: This Is Industry Restructuring, Not Industry Decline

The tech industry employed roughly 4 million software developers in the US in 2022. The AI era layoffs — while large in headline number — have not meaningfully reduced that total. What they have done is accelerated a restructuring that was going to happen over a longer period anyway.

Every major technology transition restructures the workforce that builds technology. The transition from mainframes to PCs restructured IBM. The transition to the web restructured software companies that had been selling packaged software. The transition to mobile restructured the desktop software industry. In each case, the total number of technology workers did not decline over the decade following the transition — it grew. But the specific roles, skills, and companies that led that growth changed substantially.

The AI transition is faster than prior transitions, which compresses the adjustment period. Developers who retool faster will move through the transition with less disruption. The retooling that matters most is not learning to use Cursor or Claude Code — those are table stakes. It is building the mental model of AI systems: what they can do reliably, where they fail, and how to architect products that are robust around their limitations.

Key Takeaways

• Microsoft cut 6,000 in May 2025 and kept expanding AI teams — total headcount ended approximately flat, composition changed significantly
• Meta went from 21,000 layoffs (2022–2023) to near pre-layoff headcount by 2025, rebuilt in AI research, infrastructure, and products
• Google Cloud hiring remained strong through the same period that Google consumer products contracted — the AI infrastructure side of the company is different from the threatened-product side
• Revenue per employee at Meta grew from ~$400K to ~$750K — AI productivity gains are being taken partly as cost savings, partly as capability expansion
• AI engineer roles pay 20–35% premiums over equivalent software engineer roles at the same companies — compensation is the leading indicator of where the market is moving
• This is industry restructuring, not decline — every major tech transition has produced more total tech employment over a decade, with significant role composition change

Use the Will AI Replace Me tool to assess your specific role exposure. Track where AI infrastructure investment is flowing with LLM API Pricing.