Entry level jobs are the first thing a company sells when it decides to buy a data center, and in 2026 the industry sold a record number of them. Tech employers have cut 206,877 people across 715 layoff events since January while the same handful of firms committed somewhere between $690 billion and $725 billion to AI infrastructure. The cuts did not land evenly. They landed on the work that used to be handed to a 23-year-old with a bootcamp certificate and a lot of patience: the boilerplate code, the basic testing, the ticket queue, the lead list. The executives who signed those budgets are having an excellent year. The people who were supposed to walk in behind them are outside the building with a laptop bag and a rejection email thanking them for their interest.
Two numbers explain the trade. About 121,000 American jobs were announced as AI-caused in the first ten months of 2026. About $700 billion was promised to the compute that is supposed to do the causing. The second number is a bet. The first number is a receipt.
Andy Challenger — The Money for Those Jobs Is Already Spent
Challenger, Gray & Christmas, the outplacement firm that has counted American layoffs since 1993, logged 121,401 announced U.S. cuts attributed to AI from January through October (Latin Times). That is more than double the 54,836 the firm recorded for all of 2025. AI’s share of announced layoffs climbed from roughly 5 percent to about 23 percent by the first half of this year (Challenger). The tech sector alone lost 171,501 positions through October, close to a third of every announced U.S. reduction.
May was the month the vocabulary settled. American tech companies announced 38,242 cuts in four weeks, the sector’s worst month in about two years, and AI was the leading stated cause across all employers for the fourth month running. Then the citations collapsed: by October, employers blamed AI for just 1,265 cuts. Either the technology stopped being the reason, or the reason stopped being useful.
Challenger’s chief revenue officer has been unusually direct about what his own number measures. These are employer justifications collected at the moment of announcement, not audits, and he has said the funding for the eliminated roles is gone whether or not software replaced the person. A company can be wrong about why it fired you and still be right that your seat is not coming back.
Entry Level Jobs — The Rung Everyone Assumed Would Still Be There
Boston Consulting Group’s workforce analysis splits the market three ways: roles where AI amplifies a person and demand grows, roles where it substitutes for a person outright, and roles that diverge. Software engineers sit in the first bucket. Entry-level developers sit in the third (OpenTools).
The mechanism is not exotic. The first two years of a developer’s career were never about brilliance. They were about absorbing the boring work — writing the form validation, running the regression set, cleaning up somebody else’s integration — while learning how a real system behaves when it breaks at two in the morning. Those are exactly the tasks a coding assistant now finishes in seconds and a manager no longer has to staff. Senior engineers came through the year intact. The path into senior engineering got narrower by the width of the first job.
Customer success, call centers and content moderation sit in the substituted bucket, where the swap is blunter. Gainsight cut 270 people while describing its rebuild around AI agents. The bottom of the market shows up in the boards: TrueUp counted 281,328 open roles at public tech companies and top startups in mid-November, down 41 percent from the 477,845 peak in April 2022 (TrueUp). There are still jobs. There are fewer doors, and the ones left ask for experience nobody is being allowed to acquire.
A career ladder with the bottom rung removed is not a ladder. It is a rope, and it is hanging in a building you can no longer get into.
Mark Zuckerberg — $135 Billion for Compute, 8,000 People Out the Door
Meta cut roughly 8,000 positions, about a tenth of its workforce, in May, cancelled plans for 6,000 more hires and booked more than $1 billion in severance. Support functions, non-AI product teams, integrity and cybersecurity took the heaviest losses — the departments that keep a platform from embarrassing itself — while the company guided capital spending to $135–145 billion for 2026, close to double its prior level, almost all of it for AI infrastructure (Crowdfund Insider).
An internal exercise that modeled cuts of up to 60 percent in some scenarios was scaled back after employees pushed back and executives were asked what productivity was actually being delivered. A planned second wave was cancelled. The first one was not.
Severance is the bill for a decision, not a share of the upside.
Oracle — 21,000 Jobs and One Sentence in a Securities Filing
Oracle’s headcount fell by about 21,000 over its fiscal year, roughly 13 percent of the company, and its annual filing is the rare document where a large employer puts the causal claim in writing: AI adoption, it states, has contributed to the reduction and may keep doing so, while capital spending rose 162 percent to $55.7 billion and the company raised $50 billion in debt and equity to pay for it (The Next Web). Most companies say restructuring. Oracle said the quiet part into a document that regulators read, which makes it the cleanest evidence of the year.
Amazon removed about 16,000 corporate roles in January with $200 billion of AI capital spending behind it. Microsoft cut roughly 4,800 roles in July, said those positions were not being replaced by AI, then offered voluntary exits to as many as 60,000 employees across Xbox, Windows and finance. Alphabet opened buyouts in several countries in November with $185 billion of capex committed.
An offer of voluntary exit to 60,000 people is not a restructuring. It is a hint with paperwork.
The 62% Premium — and the 17% Who Admitted the Excuse
The market that remains pays for fluency. Workers with verified AI skills earn about 62 percent more on average, AI-focused postings grow eight times faster than the rest of tech, and more than half of active IT listings now ask for some AI proficiency (Latin Times). That premium is real. It is also being collected almost entirely by people who already had a job when the year started.
The attribution underneath the headline is softer than it looks. A Resume.org survey found 17 percent of companies admitted they framed layoffs as AI-driven when finances were the actual cause, and another 42 percent conceded doing so to some degree. Gartner found no correlation between AI-linked workforce reductions and better stock returns (OpenTools). The layoff is real. The reason is sometimes marketing.
Axel’s Take — the Ledger Balances Only If You Delete the People
Run the math honestly and the story gets worse, not better. Meta’s midpoint capital budget divided by its 8,000 cuts comes to roughly $16 million per eliminated role — an illustration, not an accounting identity, because payroll savings and capex sit on different ledgers. That is the point. This was never a cost-cutting exercise. The five biggest spenders are projecting combined outlays near $700 billion against less than $35 billion in revenue from the pure-play AI vendors selling into that buildout, and free cash flow is falling at every one of them. Payroll is simply the line item that moves fastest when a board wants a data center and a story about the future.
Now price the damage the way an actuary would. Goldman Sachs analysis reported this week found that workers displaced from AI-exposed roles since 2025 take about a month longer than other laid-off workers to find new work and are roughly ten percentage points more likely to still be unemployed a year later (Crowdfund Insider). A month and ten points does not sound like a catastrophe on a spreadsheet. It is a mortgage payment, a health premium, and an eleven-month gap on a résumé that the next hiring manager reads as a character flaw.
The part nobody is pricing is 2031. You cannot staff a senior engineering bench in five years if the industry spent 2026 refusing to hire juniors. The 206,877 people counted this year are the visible cost. The invisible one is the cohort that never got a first offer — the graduates who will spend their twenties being told they lack experience by companies that deleted the job which supplies it.
Oracle put it in a filing. Meta put it in a capital budget. Microsoft put it in a buyout letter. The industry has been remarkably transparent about what it is doing. The only people still calling it a mystery are the ones being paid to.
More on this: 190 Jobs Cut by Nov 4 Because Automation Does the Work Now · 850 Jobs Cut So Qantas Can Buy the AI It Cannot Hire · 30,000 Amazon Jobs Cut, Then Recruiters Called Them Back
Sources: Latin Times, Challenger Gray & Christmas, The Next Web, OpenTools, TrueUp, Crowdfund Insider.