The AI budget is you

Oracle is planning to lay off between 20,000 and 30,000 employees. The reason isn't a downturn. Revenue is fine. The company wants to free up $8 to $10 billion to pour into AI infrastructure. That's not a rumor from an anonymous source, it's the actual strategic logic being reported: cut people, redirect their salary costs into GPU clusters and data center capacity.

Atlassian just did the same thing at smaller scale. 1,600 people, roughly 10% of the entire company, gone. The stated goal: "repositioning investment toward AI." Not because those 1,600 people were underperforming. Not because the business was struggling. Atlassian's revenue grew 15% last quarter. They cut humans because they decided the money was better spent on machines.

The math is blunt

The interesting part isn't that layoffs happen. Tech layoffs are a recurring feature of the industry. The interesting part is the stated reason. Companies used to dress layoffs in euphemisms about "organizational efficiency" or "strategic realignment." Now they're just saying it: we need the money for AI, and the money is currently being spent on you.

Oracle's math is straightforward. The company employs roughly 160,000 people. Cutting 20,000 at an average fully loaded cost of $150,000 per head frees up $3 billion annually. Cutting 30,000 frees up closer to $4.5 billion. Add in the real estate savings, the benefits overhead, the equipment budgets, and you're within range of the $8 to $10 billion target. The headcount line item on the balance sheet is being directly converted into a capex line item for AI.

Atlassian's version is similar. 1,600 employees at their compensation levels represents roughly $400 to $500 million in annual savings. That money moves straight into AI product development and infrastructure. CEO Scott Farquhar has been explicit that the company sees AI as its growth vector, and the funding has to come from somewhere.

They're not alone. Meta cut 21,000 people across 2023 and 2024, then redirected billions into AI labs and compute infrastructure. Google laid off 12,000 in January 2023, followed by smaller cuts throughout 2024 and 2025, while simultaneously announcing massive AI capital expenditure increases. The pattern is consistent enough to be a strategy, not a coincidence.

The "AI creates jobs" claim has a timing problem

Every major AI company, from OpenAI to Google to Anthropic, has published some version of the "AI will create more jobs than it destroys" talking point. The historical parallel they love is the ATM: banks installed ATMs starting in the 1970s, but the number of bank tellers actually grew because ATMs made it cheaper to open branches. The technology displaced specific tasks but expanded the overall industry.

That story might even be true in the long run. But it has a timing problem.

The ATM transition played out over 30 years. The teller workforce didn't collapse overnight. It gradually shifted, with natural attrition doing most of the work. People retired, others moved into different roles, branches hired for relationship management instead of counting cash. It was slow enough that individuals could adapt.

What Oracle and Atlassian are doing isn't a 30-year transition. It's a quarterly budget reallocation. Twenty thousand people don't get 30 years to find new roles in an AI-expanded economy. They get a severance package and a LinkedIn update.

The displacement is funding the thing causing the displacement

This is the part that should make the "AI creates jobs" crowd uncomfortable. The capital flowing into AI development isn't coming from new revenue generated by AI products. Not yet, anyway. For most companies, AI is still a cost center, a bet on future returns. The money to fund that bet is coming from existing headcount.

Oracle isn't investing $8 to $10 billion in AI because AI already earned them $8 to $10 billion. They're investing it because they believe it will. And they're funding the belief by eliminating the people who currently produce the revenue.

This creates a specific feedback loop. Companies lay off workers to fund AI. AI companies use that funding to build products that replace more workers. Those companies then lay off more people to fund more AI. At no point in this cycle does the "AI creates jobs" part kick in at scale. It's a promissory note that keeps getting extended.

The investment numbers make this concrete. Global corporate spending on AI infrastructure is expected to exceed $300 billion in 2026. A meaningful fraction of that is being funded not by new revenue but by headcount reduction. When Larry Ellison talks about Oracle's AI buildout, the source of funds isn't a mystery. It's printed on the pink slips.

Who actually gets hired

When companies say "AI creates jobs," they're technically correct in one narrow sense: the AI industry itself is hiring. Data center construction workers, ML engineers, GPU supply chain specialists, prompt engineers (for now), AI safety researchers. These roles exist and they pay well.

But the people being laid off at Oracle aren't ML engineers. They're project managers, mid-level developers, sales operations staff, technical writers, QA testers, HR coordinators. The roles AI companies are creating don't map onto the roles being eliminated. A 45-year-old program manager with 15 years at Oracle isn't going to retrain as a CUDA optimization specialist. Telling them "AI creates more jobs than it destroys" is technically an economic observation and practically useless career advice.

The honest version of the talking point would be: "AI will create a large number of high-paying jobs for people with specific technical skills, while eliminating a larger number of mid-tier knowledge work jobs. The net job count might increase eventually, but the people losing jobs and the people getting new ones are mostly different humans."

The quiet part out loud

What changed in 2026 isn't the dynamic. Companies have been automating away jobs since the invention of the loom. What changed is the honesty.

Oracle didn't say "we're restructuring for operational excellence." They said, more or less, "we need billions for AI and we're going to get it by cutting tens of thousands of positions." Atlassian didn't blame macroeconomic conditions. They pointed directly at AI investment as the reason.

This honesty is actually useful, even if it's brutal. It strips away the pretense that lets everyone feel comfortable. When a company says "strategic realignment," employees can tell themselves it might not be about them. When a company says "we're replacing your budget line with an AI budget line," there's nowhere to hide.

It also forces a more honest conversation about policy. If companies were still pretending layoffs were about efficiency, governments could pretend they didn't need to respond. When companies explicitly say "we are converting human labor budgets into AI compute budgets," the need for workforce transition programs, retraining infrastructure, and updated social safety nets becomes harder to wave away. The candor is uncomfortable, but it's better than the alternative where everyone pretends this isn't happening until it's too late to build anything to catch the people falling through.

What this means if you're watching from inside

If you work at a large technology company, the relevant question isn't whether your company is planning AI-related headcount reductions. It is. The question is when, and whether your specific role is in the first, second, or third wave.

The first wave, which is happening now, targets roles where AI can already demonstrably reduce the need for humans: content creation, basic coding tasks, first-tier customer support, QA, data entry and processing. The second wave will hit roles where AI handles the coordination and decision-support layers: project management, business analysis, parts of product management. The third wave is harder to predict, but it will likely reach into areas that feel safe right now.

The practical response isn't panic. It's positioning. The people who will survive waves two and three are the ones who are already using AI to multiply their own output, making themselves the person who manages the AI rather than the person whose work the AI replaces. That's not a guarantee, but it's better than hoping your company's CFO doesn't notice that your function can be automated.

The "AI creates jobs" narrative isn't wrong. It's just irrelevant to the person whose position got converted into a line item for Nvidia H100 purchases this quarter.

Companies are now saying openly what they're doing. The least we can do is listen.

Marco Kotrotsos writes about practical AI implementation at gloss.run and acdigest.substack.com.