GitLab Cut 14% of Staff to Fund AI Infrastructure. Visa Cut 7% Because AI Already Paid Off. The Pattern Inside Every 2026 Layoff Memo.

# GitLab Cut 14% of Staff to Fund AI Infrastructure. Visa Cut 7% Because AI Already Paid Off. The Pattern Inside Every 2026 Layoff Memo.

Published: September 14, 2026

By September 12, 2026, 2026 had recorded 365 layoff events totaling 209,032 workers, and the year is running at roughly 980 job losses a day, versus 564 a day in 2025. (skillsyncer.com) LLMHire has covered plenty of those individual cuts this year — the aggregate count in July, Cisco and Intuit cutting inside record profits, monday.com's SEC filing, Patreon's "AI isn't replacing anyone" memo. Read enough of them back to back and a more useful story emerges than "AI is killing jobs": at a growing number of companies, the layoff and the AI hiring aren't sequential. They're the same restructuring, running in both directions at once, inside the same org chart.

GitLab and Visa are the cleanest cases from the last quarter, and they show two *opposite* mechanics doing the same job.

GitLab: Cutting Humans to Buy AI Capacity

GitLab laid off roughly 350 employees — about 14% of staff — and exited 22 countries in a restructuring that cost the company $30–35 million in severance and exit charges. (TechCrunch) The company was not in distress when it did it: first-quarter revenue was $264.2 million, up 23% year over year, with 88% gross margins. (Tech Times)

What GitLab said out loud is the part worth sitting with: CEO Bill Staples framed the cut as funding "a generational rebuild" of the platform to handle AI agents committing code at machine scale — traffic that GitLab's existing infrastructure wasn't built for. The company is partnering with an outside AI lab to redesign its infrastructure around APIs built for agents to store and retrieve code context, plus new orchestration tooling to coordinate work between AI agents and human developers. (TechCrunch) This is not "AI replaced 350 people." It's "we reduced headcount specifically to redirect that budget into the infrastructure AI-driven usage now requires" — a direct trade of human payroll for AI-agent capacity, at a company whose own product is the thing agents are increasingly using.

Visa: Cutting Humans Because AI Already Worked

Visa's version runs the opposite direction. The company is cutting about 2,600 jobs — 7% of its global workforce, concentrated in technology and product teams. (CNBC) Here AI isn't the reason for the cut — it's the reason the cut is possible. CEO Ryan McInerney said new AI tooling has produced a 65% boost in feature-development speed, and Visa has restructured its product teams from roughly 10 people down to two to four people per team as a direct result. (Tech Times) The freed budget isn't sitting idle, either — Visa says it's reinvesting in stablecoin infrastructure and B2B payments growth. (PYMNTS)

Put the two next to each other and the mechanics are inverted: GitLab cut people *to afford* AI capacity it doesn't have yet. Visa cut people *because* AI capacity it already has made the old team size unnecessary. Both memos say "AI." Both are true. They are not the same sentence.

Microsoft's Version Sits in the Middle

Microsoft's July cut — 4,800 roles, about 2.1% of headcount, with roughly 1,600 of them in the Xbox division — is the third shape. (GeekWire) Microsoft's own chief people officer was explicit that none of the eliminated roles are being directly replaced by AI — the company's language was "AI is changing how work gets done," not "AI is doing this work now." (ABC News) That's neither GitLab's "buy capacity" framing nor Visa's "capacity already delivered" framing — it's closer to a hedge: shrink the org ahead of a shift the company can't yet fully quantify, while record AI infrastructure spending continues elsewhere on the balance sheet.

Oracle is the extreme end of this same spectrum — LLMHire covered its 21,000-person, 13%-of-workforce reduction over the trailing twelve months, tied directly to a nine-figure AI infrastructure buildout, in a dedicated career playbook back in the spring. What's changed by September is that Oracle is no longer the outlier. It's the first clear instance of a pattern that GitLab, Visa, and Microsoft have each now run their own version of.

The Data That Actually Answers "Is AI Killing Jobs"

The New York Fed's own survey data, as reported by TechRadar Pro, is the closest thing to a real answer: only about 4% of AI-using service firms said they'd laid off workers because of AI in the past six months, while roughly 13% said AI had caused them to hire more people. (TechRadar Pro) That's not "AI isn't affecting jobs" — it's that the net effect, at the median firm, currently skews toward AI creating hiring need faster than it creates layoffs. The headline-grabbing cuts (Oracle, GitLab, Visa, Microsoft) are concentrated at large public companies making structural, board-visible bets — not evidence that the median employer is shrinking.

What This Means If You're Job Hunting Right Now

• Read which mechanic a company is running, not just whether it cited AI. A GitLab-style cut ("we're buying AI capacity we don't have yet") means the company is actively building — infrastructure, platform, and agent-orchestration roles are the hiring side of that same memo. A Visa-style cut ("AI already made the team smaller") means the surviving team is leaner and more senior, and entry paths narrow rather than open.
• Infra and AI-ops roles are consistently the hiring side, even at companies cutting elsewhere. GitLab's own restructuring is built around engineers who can design context APIs and agent-orchestration tooling — the kind of agent/MCP tooling work covered in depth at AgenticNode — while its support and admin-heavy roles were the ones cut.
• "AI made the team faster" (Visa) is a signal to raise your own fluency bar, not avoid the company. Visa's 2–4 person product pods only work because each person is meaningfully more productive with AI-assisted tooling. That's the same dynamic behind the AI-augmented engineer salary premium LLMHire tracked in May — and it's exactly the skill gap the Vibe Coding Ebook is aimed at closing.
• Don't treat "AI" in a layoff memo as one story. Fund-the-buildout (GitLab, Oracle) and capacity-already-delivered (Visa) are different market signals wearing the same headline.

Where to Find These Roles

LLMHire tracks AI infrastructure, agent-orchestration, and AI-augmented product roles at the companies running both sides of 2026's restructuring wave — sourced from Greenhouse, Lever, Ashby, and direct company listings, updated 6× daily.

Browse AI infrastructure and agent-orchestration roles →

See AI-augmented product engineering roles →

Explore platform engineering roles for AI workloads →

Related: Record Profits, Thousands of Cuts: What Cisco and Intuit's 2026 Layoffs Reveal About Where AI Hiring Is Actually Headed · Monday.com Just Cut 620 People to Become an 'AI Work Platform.' · Patreon Cut 20% of Its Staff While Revenue Grew 28%

LLMHire tracks 6,450+ AI engineering roles from Greenhouse, Lever, Ashby, and direct company listings. Updated 6× daily. For the tooling behind AI-agent infrastructure work, see AgenticNode; for the AI coding fluency gap driving leaner product teams, see the Vibe Coding Ebook.