A year ago, every tech CEO was a prophet. They wrote the memo. They posted the tweet. They gave the fireside chat. “AI-first.” “Reflexive AI usage.” “Prove AI can’t do the job before we hire a human.” It was the performance piece of 2025 and for a hot summer the press ate every word of it.
I’m less interested in what they said. I’m interested in what happened next.
The memos were a promise. Stock charts, revenue lines, and the people who used to swipe into the building are the receipt. I pulled the numbers on every CEO who publicly declared their company AI-first — from Jensen Huang’s 2014 email to Brian Chesky’s 2025 earnings call — and lined them up from best to worst. What came back is uglier and funnier than anything I expected.
Let’s go to the scoreboard.
The winners who actually cashed the check
1. NVIDIA — Jensen Huang
Declared AI-first: 2014 (internal email)
Verdict: The only one who said it a decade early and got every dollar of what he promised.
While his peers were still selling PowerPoint decks about “digital transformation,” Jensen told his staff NVIDIA was done being a graphics card company. In the twelve years since, his revenue went from a few billion to a full-year FY26 haul of $193.7 billion, up 68% year-over-year. Q4 FY26 alone: $68.1B in revenue and $39.5B in net income — more profit in three months than most Fortune 500 companies book in a decade.
Jobs: +6,400 gained. Headcount went from roughly 29,600 to 36,000 — a 21.6% year-over-year jump. Revenue per employee around $4.1M. Profit per employee around $2.1M. Read that sentence again. He’s the only CEO on this list whose AI-first bet made the company hire more people, not fewer.
2. Alphabet / Google — Sundar Pichai
Declared AI-first: October 2016 (earnings call)
Verdict: Nine years to cook, and the search ad monster finally became the AI ad monster.
Sundar coined the phrase. Then he spent almost a decade looking like he was losing to OpenAI. Then Gemini 3 shipped, AI Overviews hit 1.5 billion monthly users, and Google Cloud ended 2025 at a $70B annual run rate with 48% Q4 growth. Full year 2025: $402.8B in revenue — Alphabet’s first-ever $400B year — and net income up 32% to $132B. Stock up about 104% year-over-year. Market cap near $3.9 trillion. Google Search alone now generates roughly $615 million a day.
Jobs: roughly 1,500–2,000 lost across scattered rounds in Android, Pixel, cloud, and HR through 2025 — modest for a workforce of ~180,000. Planned $175–185B in 2026 capex. Nobody at Alphabet is getting rich from AI productivity. They’re getting rich from selling the picks and shovels.
3. Alibaba — Eddie Wu
Declared AI-first: September 2023 (first memo as CEO)
Verdict: The quiet one. While Americans were writing LinkedIn essays, Wu was shipping Qwen.
Wu took over from Daniel Zhang in September 2023 and told staff Alibaba was “user-first and AI-driven” or it was going to be displaced. Since then: Qwen crossed 300 million monthly active users (the Qwen mobile app alone went from 31M to 203M between January and February 2026), cloud growth accelerated to 34%, and Wu committed $52 billion over three years to AI and cloud infrastructure — more than Alibaba’s entire prior decade of AI/cloud spend combined. FY2025 revenue: $137B, up 6%, with net profit up 77%. Stock up about 70% in 2025 before a healthy correction.
Jobs: net ambiguous. Alibaba’s 2023–2024 “1+6+N” restructuring trimmed thousands of roles before Wu even got to his AI pivot, but the cuts weren’t AI-justified — they were regulatory and strategic. Since the declaration, Alibaba has been hiring AI talent hard (poached a DeepMind scientist in early 2026) and Wu took direct control of the Foundation Model Task Force. This is the only company on the list where the AI bet looks like an offensive move, not a cost-cutting one.
4. Microsoft — Satya Nadella
Declared AI-first: LlamaCon, April 2025 (made it explicit)
Verdict: Printing money on Azure. Still cutting Americans.
Nadella’s claim that 20–30% of Microsoft’s code is now AI-generated turned into a quiet instruction: ship more, cost less. Azure AI workloads are growing at the rate you’d expect for the company selling OpenAI’s infrastructure. But here’s the receipt: ~9,100 employees cut in July 2025, on top of 6,000+ in May–June 2025. Xbox studios (King, Zenimax, Turn 10) got hollowed out.
Jobs: ~15,000 lost in 2025 alone. Microsoft is doing fine. Microsoft employees are the ones learning what “AI augments you” actually means when the spreadsheet runs.
5. Meta — Mark Zuckerberg
Declared AI-first: LlamaCon, April 2025 + repeated since
Verdict: Quietly the best financial performer you’re not paying attention to. About to run its biggest layoff since 2022.
2025 revenue above $200B, profit around $60B. Zuck told investors AI will write half of Meta’s code within a year. He’s shifting dollars from the metaverse-that-never-was to AI infrastructure and smart glasses. Financially, Meta is cruising.
Jobs: 5,600 lost in 2025 (3,600 in January, 500 from the AI division itself in October, 1,500 from Reality Labs in January 2026). Reuters is reporting Meta is preparing to cut roughly 8,000 more in May 2026 — with internal planning documents that contemplate up to 20% of the workforce. If the higher end happens, Meta single-handedly becomes the biggest AI-first layoff of the decade.
6. Amazon — Andy Jassy
Declared AI-first: company-wide memo, June 2025
Verdict: AWS is a cash cannon. Corporate Amazon is a funeral home.
AWS did $33B in Q3 2025, up 20% year-over-year. Amazon announced $100B+ in AI/cloud infrastructure spend through 2026. Jassy’s June memo said out loud what everybody already knew: “We will need fewer people doing some of the jobs that are being done today.”
Jobs: 30,000 corporate employees lost between October 2025 and January 2026 (14,000 + 16,000 in two waves), roughly 10% of the 350,000-person corporate headcount. Jassy later told analysts it was “culture, not AI.” Sure, Andy. The memo from June said one thing, the pink slips in October said the same thing, but the earnings call said “culture.” We can all read.
The middle — revenue up, stock sideways, nobody celebrating
7. Airbnb — Brian Chesky
Declared AI-first: August 2025 Q2 earnings call
Verdict: Said it late. Shipped it well. Market yawned.
Chesky was the last to the AI-first altar and the first to do it without firing anybody. 2025 revenue: $12.2B (+10%). Q4 “Nights and Experiences” bookings up 10% to 121.9 million — the best growth quarter in over a year. FCF of $4.6B. He hired Ahmad Al-Dahle (ex-Meta, ran Llama) as CTO in January 2026. AI-powered natural language search — “search by vibe” — actually increased conversion rates.
Jobs: flat. No mass layoffs announced. Stock down about 3% YTD 2026. The market is rewarding him with indifference, which on this list is a win.
8. Citigroup — Jane Fraser
Declared AI-first: mandatory AI training for 175,000 employees, 2024–2025
Verdict: Best-performing big bank stock of 2025. On a path to fire a city.
Stock up 67% in 2025 — #1 among major US banks. Fraser was named Euromoney Banker of the Year. She’s also on track to eliminate 20,000 jobs by the end of 2026 as part of a broader transformation, with overall headcount going from 240,000 (2022) to a target of 180,000 by end of 2026.
Jobs: ~15,000 lost so far, another 5,000+ coming in March 2026. Much of this is restructuring and the Mexico retail spin-off, not pure AI — but Fraser explicitly cites “efficiencies we have gained through technology.” When the CEO writes that sentence, HR hears “Thursday.”
9. Shopify — Tobi Lütke
Declared AI-first: the memo heard ‘round the world, April 7, 2025
Verdict: The business is fantastic. The stock is cratering. The employees are exhausted.
Revenue +30% for 2025 to $11.6B. Q4 +31%. Free cash flow $2B. Ten straight quarters of double-digit FCF margins. A $2B share buyback. By every financial metric Lütke is winning. By every other metric — down 28–30% year-to-date in 2026, market cap drifted from ~$250B CAD to ~$144B USD, multiple rounds of layoffs through late 2025 and January 2026. The Logic ran a whole feature on how Shopify employees keep lists of friends’ Slack profiles to watch for the grey-out.
Jobs: multiple rounds of layoffs in November 2025 and January 2026 on top of the famous partnerships cuts. Total headcount around 8,100. Hard exact count on AI-linked firings, but the pattern Lütke set — “prove AI can’t do it” — produces firings by design. Call it ~1,000+ in the year since the memo, conservatively.
The danger zone — stock punishment, people pain
10. Salesforce — Marc Benioff
Declared AI-first: Agentforce launch, Dreamforce 2024; ramped up 2025
Verdict: Agentforce is selling. The stock doesn’t care. Four thousand support engineers don’t care either.
Agentforce hit $1.4B ARR with 9,500 paid deals by Q3 FY26, a 114% year-over-year gain. Revenue guidance raised to $41.5B. So why did the stock fall about 45% in 2025? Because enterprise SaaS is the category most exposed to agentic disruption — investors think the same Agentforce Marc is selling could eat Salesforce’s own seat-based licensing model alive.
Jobs: 4,000 customer service roles eliminated. Benioff on a podcast: “I’ve reduced it from 9,000 heads to about 5,000, because I need less heads.” In a separate interview he called AI “the lazy way out” for other CEOs cutting jobs. Meanwhile his own support team is half the size it was.
11. Box — Aaron Levie
Declared AI-first: internal memo, May 1, 2025
Verdict: Said the nice version. Got the nice result. Nobody cares.
Levie’s memo was the “AI as ladder” version — use AI well and you get more headcount. The business: Q3 FY26 revenue $301M (+9%). Enterprise Advanced now 10% of revenue after one year. FY26 total revenue ~$1.17B. Stock: trading around $23–$24, basically flat to slightly down, below analyst targets.
Jobs: no announced layoffs. For an established SaaS company, Box is arguably the model for how to do this without crushing your own workforce. Investors are yawning. Employees still have badges.
The wreckage
12. Klarna — Sebastian Siemiatkowski
Declared AI-first: sustained campaign 2023–2025, including the AI avatar earnings call
Verdict: Went further than anyone. Reversed course on live television. Then IPO’d at a third of the peak valuation.
Siemiatkowski shrunk headcount from 5,527 (Dec 2022) to 3,422 (Dec 2024) — about 2,100 jobs eliminated. Bragged the AI chatbot did the work of 700 customer service agents. Recorded an AI avatar to present quarterly earnings. Then in May 2025 he told Bloomberg the whole thing produced “lower quality” and Klarna was hiring humans again — Uber-style gig workers at ~$41/hour. The September 2025 IPO priced at $40/share for a $17B valuation — down from Klarna’s 2021 peak of $45.6 billion. First quarterly earnings after IPO: a loss.
Jobs: ~2,100 net lost. Quietly rehiring. The most honest failure on this list, which somehow makes it the least humiliating.
13. Duolingo — Luis von Ahn
Declared AI-first: memo, April 2025
Verdict: Stock off a cliff. Backtracked the policy. Apologized on the New York Times.
The green owl memo produced the biggest public backlash of the entire AI-first wave — “AI first means people last” trended. Duolingo crossed $1B in bookings for the first time in 2025, DAU hit 52.7M (+30%). And the stock collapsed: from $321.84 in September 2025 to $175.50 by December 2025, and then fell further after weak 2026 guidance in February 2026. Von Ahn publicly admitted to the NYT that he “did not give enough context,” confirmed zero full-time employees were laid off, and a year later told the Silicon Valley Girl podcast that Duolingo no longer considers AI use in performance reviews. He folded the whole memo.
Jobs: 0 full-time lost. Contractor program shrunk and contractor hiring paused in areas AI can cover. The saving grace for von Ahn is that he didn’t actually fire humans for his slogan. The cost is that the slogan cratered his stock anyway.
14. Fiverr — Micha Kaufman
Declared AI-first: “AI is coming for your jobs” memo, April 2025
Verdict: Was right about AI. Got eaten by it.
Here’s the grim irony. Kaufman’s whole business — a marketplace where small businesses hire freelancers to design logos, write copy, edit video — is exactly the work AI is automating first. In September 2025 he did his own AI-first restructuring: laid off 250 people, about 30% of Fiverr. For 2025 he squeaked out 10.1% revenue growth to $430.9M. Then the 2026 guidance dropped: revenue expected to DECLINE to $380–420M — a 3% to 12% drop even in the optimistic case. Stock fell 35%+ YTD 2026. Market cap around $445M — barely above the cash the company is sitting on — which means the market is pricing the operating business at effectively zero.
Jobs: 250 lost at Fiverr. Plus countless freelancers whose gigs got replaced by a prompt.
15. IgniteTech — Eric Vaughan
Declared AI-first: ongoing through 2024–2025
Verdict: The extreme version, at small scale.
Vaughan ran a 30-year-old enterprise software company, gave every employee $1,200 to take AI courses — and fired 80% of his staff for not adapting fast enough. He told Fortune he’d do it again. At his scale (small, private) the absolute numbers are modest, but the percentage is the most aggressive on this list. Proof of concept for the founders watching.
The tally
Let’s count the bodies.
Confirmed job losses at AI-first declared companies since the memo wave:
Amazon: 30,000 corporate (Oct 2025 – Jan 2026)
Citigroup: ~15,000 so far, on a path to 20,000 by end of 2026
Microsoft: ~15,000 (May – July 2025 combined)
Meta: 5,600 confirmed — plus ~8,000 more planned for May 2026, with up to 16,000 on the table
Salesforce: 4,000 (customer support roles)
Klarna: ~2,100 net (now partially rehiring as gig workers)
Alphabet / Google: ~1,500–2,000 scattered rounds
Shopify: ~1,000+ across multiple rounds
Fiverr: 250 (30% of workforce)
IgniteTech: 80% of staff (small base)
Duolingo: 0 FTE — but contractors pruned
Box: 0
Airbnb: 0
Alibaba: strategic restructuring predates the AI pivot; hiring AI talent aggressively
NVIDIA: +6,400 HIRED
Running total of confirmed AI-era job losses at these companies alone: roughly 74,500.
Jobs gained at these companies (essentially one name, NVIDIA): about 6,400.
Net: a shortfall of around 68,000 jobs — and that’s the conservative number, before Meta’s May 2026 cut lands, before Citi’s March 2026 round, before whatever Jassy calls “culture” next quarter. With the committed-but-unexecuted cuts at Meta and Citi, we’re staring down 90,000+ jobs erased at this slice of the economy before the first anniversary of most of these memos.
Zoom out and the broader number is worse. InformationWeek tracked roughly 245,000 tech jobs eliminated in 2025 globally, with about 55,000 explicitly attributed to AI. Over a million tech jobs gone since 2021. That’s the context the memos are sitting inside.
Close
Here’s the part most people still won’t say out loud. The CEOs who wrote the memos were not wrong. They were clumsy, self-serving, often cruel about it — but they were reading the room. NVIDIA is the most valuable company on earth because it sells the tools. Alphabet just booked a $400 billion year because AI is driving search usage up not down. Alibaba is building a token economy. Meta is hiring DeepMind researchers while firing Reality Labs managers. The winners are not the companies that used AI to shrink. The winners are the companies that used AI to build. The losers are the ones who mistook a cost-cutting memo for a strategy.
The real signal isn’t in the memos. It’s in the scoreboard. Fiverr’s entire operating business is priced at zero. Think about that for thirty seconds. A public company’s market cap is barely above its cash pile because investors have concluded the work Fiverr sells can now be done by a model for a dollar. A year ago Fiverr’s CEO told his staff AI was coming for their jobs. He was right. He just didn’t realize it was coming for his company too.
Duolingo crossed $1 billion in bookings and the stock still collapsed. Klarna IPO’d at a third of its peak valuation after being the loudest AI-first voice in fintech. Salesforce — the company selling the agents — lost 45% of its market value in the year it went hardest on Agentforce. These aren’t opinions. These are the closing prices.
Seventy thousand people at these companies got a severance letter in the year since the memos started dropping. Another twenty to thirty thousand are scheduled for 2026. One company — one — hired more humans. The rest turned their employees into a line item.
The world has already changed. The memos were just the CEO class catching up to what every engineer running a local model already knew. The scoreboard is now public, the bodies are counted, and the stock charts are writing the ending in real time.
If you still can’t see it, you’re not skeptical. You’re blind.