It's always surprising when an employee comes to me and they make an argument that the company is making much more per employee than what it pays to employees. The thing is that it's just a normal thing for any profitable business, but now AI is increasing that ratio at an incredible rate.

Two days ago Microsoft offered voluntary exit packages to about 60,000 people, roughly a quarter of its workforce12. It wasn't a company in trouble. Three months earlier it had reported its best year ever: $331.8 billion in revenue, up 18%, with operating income up 21%3 and Azure passing $100 billion for the year, up 41%4.

For most of my career, big layoffs meant something had gone wrong: revenue was falling, margins were shrinking, the company was cutting to survive. Microsoft is cutting while it grows. 

So does AI actually raise revenue per employee, and if it does, who gets the difference?

Twenty years of Microsoft in three numbers

In fiscal 2006 Microsoft had about 71,000 employees and $44 billion of revenue5. By 2020 it had about 163,000 people and $143 billion6. At the end of June this year it had 223,000 people, its first headcount decline in a decade, and $332 billion of revenue7.

Index both lines to 2006 and the gap opens up. Revenue grew about 7.5 times. Headcount grew about 3.1 times.

Microsoft revenue grew more than twice as fast as its workforce

Fiscal years, indexed to 2006 = 1

RevenueHeadcount

Times the 2006 level

Source: Microsoft 10-K filings and earnings releases

Divide one by the other and you get revenue per employee: about $620,000 in 2006, $880,000 in 2020 and almost $1.5 million in 2026. Now run the thought experiment the buyout invites. If all 60,000 people took the offer and nobody replaced them, Microsoft would be back to its 2020 headcount while making well over twice its 2020 revenue. Even at this year's revenue, that's more than $2 million per person.

That won't happen in full. A voluntary offer isn't 60,000 departures, and this is Microsoft's second programme this year after a smaller one in April8. 

The rest of big tech is already there

Microsoft isn't even the high end. Using each company's latest annual report:

Revenue per employee, latest fiscal year

Revenue divided by year-end headcount, in millions of dollars. Lovable uses annualised revenue.

Big techAI-native

Source: Company 10-K filings; TechCrunch for Lovable

Nvidia makes about $5 million per employee9. Meta, Apple and Alphabet are all above $2 million101112. Amazon looks low only because most of its 1.6 million people work in warehouses13.

The newest companies make the point more sharply. Lovable crossed $400 million in annualised revenue in February with 146 full-time employees, about $2.7 million each14. Annualised revenue isn't the same as booked revenue, and young companies grow into their headcount later. 

Ten years ago you needed thousands of people to get to that revenue. Now you need a team that fits in one office.

The new pattern: grow, then shrink the team

What's changed isn't that companies want to be efficient. They always did. What's changed is that leaders now say out loud that AI is the reason they need fewer people:

CompanyWhat happened
SalesforceMarc Benioff said AI agents let him cut support from about 9,000 people to about 5,00015
ShopifyTobi Lütke's memo: before asking for headcount, teams must show why AI can't do the job16
KlarnaRevenue per employee rose to nearly $1 million as headcount fell and revenue grew17
Duolingo"AI-first": contractors phased out where AI can do the work, full-time staff kept18
Microsoft60,000 voluntary exit offers in a record year1

Some of this is a story companies like to tell investors. Not every cut is caused by AI, and some "AI layoffs" are ordinary cost-cutting with better branding. But the hiring side is harder to fake. When the default becomes "prove a model can't do it first", headcount stops growing with revenue.

So why isn't it in the statistics?

At the level of individual companies, revenue per employee is clearly rising. At the level of the whole economy, it's much harder to see.

US productivity grew 2.2% in the year to the second quarter of 2026, and it has averaged about 2.1% a year since the end of 201919. That's better than the 2010s, but it isn't a clear break. Mary Daly of the San Francisco Fed put it plainly earlier this year: most studies find limited evidence of a significant AI effect on productivity so far, and "we are simply not there yet"20. It's Robert Solow's old line about computers, updated: you can see AI everywhere except in the productivity statistics.

There are good reasons for the gap. Revenue per employee is a crude measure. It goes up when a company outsources work, raises prices, changes its business mix or spends heavily on machines instead of people. Microsoft spent more than $145 billion on capital expenditure in fiscal 202621. Some of its revenue per employee is really revenue per data center.

And the gains that do exist aren't reaching paychecks yet. A large study of Danish workers found that AI chatbots had almost no effect on earnings or hours, and that only a small share of the time saved turned into higher wages22. Stanford researchers found something sharper in the US: employment for 22 to 25-year-olds in the jobs most exposed to AI fell 13% relative to others, while older workers in the same jobs kept growing23. 

Capital wins, labour waits

What's interesting is that the share of US business income that goes to workers fell to 52.8% in the second quarter of 2026, the lowest since the series began in 194724. For most of the post-war period it was somewhere between 60% and 66%.

Where the money goes

52.8%

US labour share of business income

Q2 2026, lowest since 1947

$1.49M

Microsoft revenue per employee

+140%

vs FY2006

2.2%

US productivity growth

year to Q2 2026

That's what higher revenue per employee looks like from the other side. If companies grow revenue faster than they grow payroll, the difference goes to whoever owns the company. For shareholders it's extraordinary operating leverage. For everyone else it's a question about whether working hard still leads somewhere.

Which is why I think "affordability" became the word of this year. It isn't just about prices. It's about whether people believe they can still take part in the progress around them. A recent French survey shows how much that depends on what you already own: 58% of people with no assets named purchasing power as a top concern, against 33% of people with more than €300,00025. Same economy, two very different experiences. One group owns a piece of the growth. The other has to save faster than asset prices rise.

AI adds a second layer to that. Not just "can I afford things today", but "will the career I'm planning on still exist". 

People take on mortgages and long-term commitments because they expect their income to grow. Shrinking corporate ladders make that expectation harder to hold.

So, does AI raise revenue per employee?

At the level of companies, yes, and fast. Microsoft, Klarna and a new generation of tiny AI-native companies show that revenue can grow far faster than headcount. At the level of the economy, not yet in a way the statistics can prove, and the gains so far are going to capital, not wages.

The real issue is time. A company can reorganise in a quarter. A person may need years to retrain, and a government has to keep paying its bills the whole way through. The great paradox of AI may be that the economy gets far more productive while the usual ways people share in that productivity get less reliable. 

Keeping those paths open, through ownership, skills and the first rungs of the career ladder, might matter more over the next few years than any productivity statistic.

References 27

  1. BNN Bloomberg, "Microsoft offers voluntary exit packages to 60,000 staff" (26 Oct 2026), https://www.bnnbloomberg.ca/business/company-news/2026/10/26/microsoft-offers-voluntary-exit-packages-to-60000-staff/ ↩
  2. GeekWire, "Microsoft makes new 'voluntary exit' offer to about a quarter of its workforce, including some execs" (Oct 2026), https://www.geekwire.com/2026/microsoft-makes-new-voluntary-exit-offer-to-more-than-a-quarter-of-its-workforce-including-some-execs/ ↩
  3. Microsoft, "FY26 Q4 earnings press release" (29 Jul 2026), https://www.microsoft.com/en-us/investor/earnings/fy-2026-q4/press-release-webcast ↩
  4. Fortune, on Azure passing $100 billion in annual revenue (29 Jul 2026), https://fortune.com/2026/07/29/microsoft-azure-100-billion-annual-revenue-earnings-revenue-cloud-ai/ ↩
  5. Redmond Magazine, "Where Did Microsoft Put 10,000 New Employees?" (30 Aug 2006), https://redmondmag.com/articles/2006/08/30/where-did-microsoft-put-10000-new-employees.aspx ; Microsoft FY2006 10-K, https://www.sec.gov/Archives/edgar/data/0000789019/000119312506180008/d10k.htm ↩
Show all 27 sourcesShow fewer
  1. Microsoft FY2020 10-K, https://www.sec.gov/Archives/edgar/data/789019/000156459020034944/msft-10k_20200630.htm ↩
  2. Microsoft FY2026 10-K, https://www.sec.gov/Archives/edgar/data/0000789019/000119312526323660/msft-20260630.htm ↩
  3. Engadget, "Microsoft is reportedly offering voluntary buyouts to up to 7 percent of its employees" (Apr 2026), https://www.engadget.com/big-tech/microsoft-is-reportedly-offering-voluntary-buyouts-to-up-to-7-percent-of-its-employees-200050484.html ↩
  4. Nvidia FY2026 10-K, https://www.sec.gov/Archives/edgar/data/1045810/000104581026000021/nvda-20260125.htm ↩
  5. Meta 2025 10-K, https://www.sec.gov/Archives/edgar/data/1326801/000162828026003942/meta-20251231.htm ↩
  6. Apple FY2025 10-K, https://www.sec.gov/Archives/edgar/data/320193/000032019325000079/aapl-20250927.htm ↩
  7. Alphabet 2025 10-K, https://www.sec.gov/Archives/edgar/data/1652044/000165204426000018/goog-20251231.htm ↩
  8. Amazon 2025 10-K, https://www.sec.gov/Archives/edgar/data/1018724/000101872426000004/amzn-20251231.htm ↩
  9. TechCrunch, "Lovable says it added $100M in revenue last month alone, with just 146 employees" (Mar 2026), https://techcrunch.com/?p=3101477 ↩
  10. Outlook Business, on Marc Benioff cutting support roles with AI agents (Aug 2025), https://www.outlookbusiness.com/artificial-intelligence/ai-agents-replaced-4000-customer-support-jobs-at-salesforce-says-ceo-marc-benioff ↩
  11. Fortune, on Shopify's "AI usage is now a baseline expectation" memo (8 Apr 2025), https://fortune.com/2025/04/08/shopify-ceo-ai-automation-no-new-hires-tech-jobs/ ↩
  12. TechCrunch, "Klarna's revenue per employee soars to nearly $1M thanks to AI efficiency push" (19 May 2025), https://techcrunch.com/2025/05/19/klarnas-revenue-per-employee-soars-to-nearly-1m-thanks-to-ai-efficiency-push/ ↩
  13. Fortune, on Duolingo's AI-first memo (18 Aug 2025), https://fortune.com/2025/08/18/duolingo-ceo-admits-controversial-ai-memo-did-not-give-enough-context-insists-company-never-laid-off-full-time-employees ↩
  14. U.S. Bureau of Labor Statistics, "Productivity up 2.2 percent from second quarter 2025 to second quarter 2026" (2026), https://www.bls.gov/opub/ted/2026/productivity-up-2-2-percent-from-second-quarter-2025-to-second-quarter-2026.htm ↩
  15. Federal Reserve Bank of San Francisco, Mary C. Daly, "The AI Moment? Possibilities, Productivity, and Policy", Economic Letter 2026-06 (23 Feb 2026), https://www.frbsf.org/?p=263220 ↩
  16. Constellation Research, on Microsoft's fiscal 2026 results and capital spending (Jul 2026), https://www.constellationr.com/insights/news/microsoft-azure-q4-revenue-surges-41-tops-100-billion-annual-revenue ↩
  17. Anders Humlum and Emilie Vestergaard, "Large Language Models, Small Labor Market Effects", NBER Working Paper 33777 (2025), https://ideas.repec.org/p/nbr/nberwo/33777.html ↩
  18. Erik Brynjolfsson, Bharat Chandar and Ruyu Chen, "Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence", Stanford Digital Economy Lab (26 Aug 2025), https://siepr.stanford.edu/publications/working-paper/canaries-coal-mine-six-facts-about-recent-employment-effects-artificial ↩
  19. U.S. Bureau of Labor Statistics, "Labor share at its lowest level, 52.8 percent, in second quarter 2026" (15 Sep 2026), https://www.bls.gov/opub/ted/2026/labor-share-at-its-lowest-level-52-8-percent-in-second-quarter-2026.htm ↩
  20. L'ObSoCo and Sciences Po CEVIPOF, "Baromètre des priorités françaises", wave 5 (Oct 2026), https://www.sciencespo.fr/cevipof/fr/etudes-enquetes/barometre-priorites-francaises-l-ob-so-co-cevipof/ ↩
  21. Congressional Budget Office, "The Budget and Economic Outlook: 2026 to 2036" (Feb 2026), https://www.cbo.gov/publication/61882
  22. Insee, Informations Rapides no. 239 on public debt at the end of Q2 2026 (29 Sep 2026), https://www.insee.fr/fr/statistiques/9053525