Snap had its own moment last Wednesday. The company announced it is cutting 16% of its global workforce — roughly 1,000 full-time employees — and closing at least 300 open roles on top of that. The stock jumped as much as 11% in pre-market trading. By midday the market had decided this was, on balance, good news.
And the stated reason for the cuts, for once, was not buried three paragraphs into the 8-K. CEO Evan Spiegel put it on the first page: more than 65% of new code at Snap is now being generated — or substantially written — by AI tools. That is the sentence that paid for the stock’s 11% pop.
The numbers
Per Spiegel’s internal memo and the subsequent CNBC and Deadline reporting:
- ~1,000 full-time roles eliminated — 16% of the global headcount
- 300+ open positions closed — never being filled
- US severance: 4 months of pay, plus healthcare continuation, equity vesting, and career-transition support
- Annualized cost savings target: >$500 million by the second half of 2026
- Stock move: +11% pre-market, settled at about +9% by the close — one of Snap’s better trading days of the year
Snap’s own disclosures (via GeekWire) show 95 of those cuts are in Washington state alone, mostly from engineering and operational functions. It’s a Santa Monica company, but the pain is distributed.
The 65% number is the story
Spiegel’s memo described the layoffs as a “necessary evolution” driven by the “rapid integration of generative AI across software development and administrative pipelines.” The specific claim he put in writing — via FinancialContent’s coverage of the internal memo — is that AI now produces or substantially assists more than 65% of new software code written inside Snap.
Two years ago, “AI wrote our code” was a demo-day line. This week, it was the justification line in an S-1-style severance announcement that cut 1,000 people. The gap between those two sentences, in calendar time, is about 20 months.
For context: Microsoft CEO Satya Nadella said in early 2025 that AI was generating “20% to 30%” of code at Microsoft. GitHub’s own data from 2025 put Copilot suggestions at ~46% acceptance among active developers. Snap’s number — a full-company, whole-pipeline figure of 65% and climbing — is the highest public admission we have seen from any publicly-listed software company to date. It is not an aspirational roadmap figure. It is what Snap just told its now-smaller engineering org it is doing now.
The activist investor footnote
The bit you have to squint to see is that Snap was already under pressure from Irenic Capital, an activist hedge fund that has spent the last quarter pushing Snap toward “a clearer path to GAAP profitability” — the kind of demand that, in 2026-speak, translates to fire people and tell the market it was the robots’ idea. Spiegel did not name Irenic in the memo. He did not need to. The $500M annualized run-rate savings is, more or less exactly, the delta between Snap’s current cost structure and the GAAP-profitability line Irenic has been asking for in slide decks to the board.
This is the second useful thing to notice about this layoff cycle: AI is not always the cause. Sometimes AI is the cover story an activist investor was already demanding anyway. The one that is most flattering to the stock price gets printed in the memo.
Where the money is going
Spiegel’s memo is also clear that this is a reallocation, not a pure retrenchment. Savings are being redirected into Snap’s AR hardware division — i.e., Spectacles, the smart-glasses bet Snap has been quietly running for nine years while Meta and Apple dominate the headlines. The implicit math: each engineer Snap keeps, augmented by AI tooling, is now worth more than 1.6 engineers from the 2024 baseline, so the freed-up compensation dollars go to hardware. The engineering job isn’t gone from the balance sheet — it just moved from “write the app” to “build the glasses that wear the app.”
Whether Spectacles becomes a credible third AR platform is a separate bet. What is no longer debated is the mechanism: Snap just monetized its own engineers’ productivity gains directly, by subtracting a thousand of them.
What to watch next
- Meta’s 8,000-person cut in May, announced April 18, is the obvious follow-on event. Spiegel is not the only social-media CEO with AI-generated-code slides this quarter; he is just the one who said the number out loud first.
- The 65% figure is now a reference point. Every other public software company will be asked its equivalent number on the next earnings call. Expect a wave of disclosures in the 45–75% range over the next two quarters, which will be accompanied — surprising no one — by correlated headcount adjustments.
- The activist-investor pattern matters. When an “AI efficiency” memo coincides suspiciously neatly with an activist’s public profitability target, the AI is doing a lot of rhetorical work. That doesn’t make the layoffs fake — it makes the AI attribution part of them a marketing choice.
Snap just told the market two things: that more than 65% of its new code is being written by AI, and that it no longer needs 1,000 of the people who used to write it. The market applauded both claims, in the same trade. The next CEO reading this transcript at their own board meeting is taking notes.