In one calendar week — Tuesday April 28 through Thursday April 30, 2026 — the four largest software companies in the world reported Q1 results, raised their 2026 capex guides, and laid off (or asked to be laid off) more than 23,000 of their own employees. The Washington Post writeup that landed this morning counted how many times CEOs said the word “efficiency” on those calls.
Fifteen.
Sam Altman, in a separate appearance referenced by the Post, has a name for what’s happening: AI washing. Companies attributing to AI a workforce decision they had already made, because saying “we hired too many people during ZIRP and the music stopped” sounds worse than saying “AI is letting us reallocate.”
The disconnect is clean enough to graph.
The numbers, side by side
Earnings side, per Bloomberg’s earnings bonanza recap, Silicon Republic, and The Motley Fool:
- Meta: Q1 revenue $56.3B, up 33% YoY. Net income $26.8B. 2026 capex revised to $125B–$145B, up from a prior $115B–$135B.
- Microsoft: Azure +40% YoY, AI revenue run-rate north of $37B annualized, capex on track for the upper end of $80B fiscal-year guide.
- Alphabet: Cloud +35%, Search ad revenue resilient, capex stepped up.
- Amazon: AWS roughly flat-to-positive against tougher comps; full-year capex commentary pointed at the same higher band.
Combined Q1 revenue across the four: ~$430B. Combined 2026 capex commitments now sit at approximately $650B — the largest single-year private-sector capex pool in the history of capitalism. Per Yahoo Finance, that’s a hair above the entire 2024 GDP of Sweden.
Headcount side, per Yahoo Finance’s coordinated-cuts piece, Axios, and The Next Web:
- Meta: 8,000 confirmed layoffs starting May 20, plus 6,000 open requisitions cancelled — effective headcount reduction ~14,000.
- Microsoft: voluntary separation program offered to ~7% of US staff (~8,750 workers) — the first companywide buyout in Microsoft’s 51-year history.
- Amazon: ~30,000 cumulative since October 2025 with no fresh wave this week, though the new round of corporate cuts is still rippling through divisions.
- Alphabet: quieter — surgical cuts in Cloud sales and DeepMind operations, not the headline.
Tally just the fresh material from this week: 8,000 + 6,000 + 8,750 = 22,750. Add the residual Amazon trickle and you clear 23,000.
Why “efficiency” appeared 15 times
Read any of the four earnings calls and a verbal pattern jumps out. It is not “we are firing people because AI does their job.” It is, almost word-for-word at all four companies:
“We’re driving efficiency to fund the infrastructure investments needed to capture the long-term opportunity.”
That’s the Meta variant. Microsoft’s was almost identical. Amazon used the word “discipline” instead of “efficiency” but the grammar is the same: a passive-construction explanation that puts the blame on macro circumstance, not on a decision the executive team made.
Per the Post’s reporting:
Across earnings calls from Meta, Microsoft, Alphabet, and Amazon this week, the word “efficiency” was uttered 15 times — never directly tied to a specific role or function being automated.
This is the rhetorical move Altman flagged. If a layoff were truly caused by AI doing the work, you would expect a CFO to say so directly, because that frames the cut as growth-positive, not contraction-coded. They are choosing not to. The reason they are choosing not to is that the actual cause is closer to: we over-hired in 2021–2022 at near-zero rates, the cost of capital quintupled, AI-infrastructure capex now eats every dollar of operating leverage we used to spend on headcount, and laying off 8,000 people is the only line item that flexes faster than data-center construction.
The math under the rhetoric
The simplest way to see the picture is to put capex and headcount on the same table.
| Company | 2026 Capex Guide | Q1 Revenue | Fresh Layoff Wave (this week) |
|---|---|---|---|
| Meta | $125B–$145B | $56.3B | 8,000 + 6,000 reqs cancelled |
| Microsoft | ~$80B+ | (combined seg) | 8,750 voluntary buyouts |
| Alphabet | ~$75B | (combined seg) | surgical only |
| Amazon | ~$105B | (AWS resilient) | ~30K cumulative since Oct ‘25 |
The capex line is what is structurally rising. Per the siliconrepublic recap, Meta CFO Susan Li attributed the mid-cycle capex bump specifically to higher GPU and component pricing — i.e. NVIDIA, TSMC, and the memory consortium are extracting margin from the hyperscalers, and the hyperscalers are extracting it back from headcount.
Headcount cuts, in this framing, are not paying for AI capacity. They are paying for AI capex inflation. The two sound similar; they are not. One is a productivity story. The other is a margin-compression story dressed up as a productivity story.
Meta’s stock dropped roughly 6% after-hours on the capex revision — proof that investors heard the second story under the first one. Alphabet and Amazon, with cleaner Cloud-margin trajectories and less aggressive headcount theater, were rewarded.
Where this puts LostJobs’s running narrative
Three thoughts.
One — the “AI is replacing me” framing is sometimes literally true and sometimes a coat of paint. Snap’s October 2025 layoff memo was specific: “65% of code shipped at Snap was written by AI in the last sprint.” That is a substantive AI-attribution claim and it deserves the AI-attribution headline. Meta’s 8,000-job memo this week says efficiency. That is a budget-attribution claim wearing an AI-attribution costume. LostJobs should keep the distinction visible.
Two — the Q1 numbers vindicate Levie and Slok last week. Aaron Levie’s argument was that the 100,000 tech layoffs of 2026 are mostly a re-pricing of pandemic-era over-hire, and that the wave will not generalize to non-tech corporates because their cost structure and AI-adoption maturity are different. The earnings call language confirms this. None of the four CFOs said “AI replaced them.” They said “efficiency.” Efficiency is the language of budget, not automation. That maps to a tech-only wave, not an economy-wide one.
Three — the deception is asymmetric and short-lived. A CEO who blames AI for a layoff today gets to be on the cutting edge of the most-discussed economic story of the decade. The same CEO, in two years, will be answering shareholder questions about why the headcount reduction did not show up in operating leverage. If AI did 8,000 people’s worth of work, the residual ~70,000 employees should be doing 8.5% more output per head by Q3. If they aren’t — and the Q3 print is the test — the “efficiency” line collapses into the older line: we over-hired. Which is a less flattering thing to say in 2026 than it was in 2023, because it begs the question of why nobody figured it out faster.
The dry coda
Nobody in 2024 thought the line item that most strongly correlated with hyperscaler capex would be the corporate-buyout offer. Microsoft has not done a voluntary separation program in 51 years. It is doing one now. It is doing one in the same week it announced its biggest single-quarter AI revenue print ever.
The “AI washing” framing is correct, but it underclaims. The honest sentence is something like: we are paying NVIDIA and TSMC out of the same pocket we used to pay 22,750 people, and we would prefer the public conversation about that be about AI productivity rather than about cost-of-capital.
That sentence does not make a good earnings-call talking point. So we get fifteen “efficiencies” instead.
The Post does not award a Pinocchio. LostJobs would.