Big Tech: Revenue & profit per employee, 2018–2024
Productivity tells a different story than headcount alone. Toggle companies, switch between revenue and net income, or index everything to 2018 = 100 for a normalized comparison.
* Amazon corporate only. Amazon’s total headcount of ~1.56M includes roughly 1.2M warehouse, fulfillment, and delivery workers. This chart uses estimated corporate / tech headcount (~175K in 2018 rising to a peak of ~415K in 2022, then ~350K in 2024), per The Information’s reporting on Amazon’s five-year corporate hiring binge. Stripped of the logistics workforce, Amazon’s per-employee productivity is directly comparable to its software-first peers.
Sources: SEC 10-K filings and company earnings releases. Employee counts are fiscal year-end (annual). Microsoft fiscal year ends June; FY2018 net income was depressed by a $13.7B one-time TCJA tax charge. Apple fiscal year ends September. Amazon FY2022 net income was negative (–$2.7B) due to Rivian mark-to-market losses. Dashed vertical lines mark key market inflection points.
The productivity data points to something structural
A confident narrative has been circulating recently is that Big Tech over-hired after the pandemic, and the layoffs since 2022 are a healthy correction back to fighting weight. After hearing it for the hundredth time in one of the podcasts I was listening to this weekend, I was bothered by the same thing that has bothered me every time the argument comes up.
Without measuring revenue per employee and profit per employee, "they hired too many people" is a headcount observation, not an analysis.
So I spent some time this weekend digging through the earnings reports, the academic research, and the macro data to run it (thank you, Claude). Where I landed is somewhere most of the commentary misses: these companies are not primarily victims of their own hiring decisions, but manifestations of a stock market that runs on narrative, and the pendulum swings are a structural feature of that relationship.
Start with the productivity test. Across the companies at the center of this debate, revenue per employee is meaningfully higher in 2025 than it was in 2018. Apple went from roughly $2.0M to $2.5M per employee, Meta from $1.6M to $2.55M, Alphabet from $1.4M to $2.1M, Microsoft from $843K to $1.24M. Amazon, once you strip out the warehouse workforce that distorts the comparison, sits at roughly $2.0M per corporate employee, up from $1.3M. The layoffs did not return these companies to their pre-pandemic baseline. They left them operating well ahead of it.
The layoffs did not return these companies to their pre-pandemic baseline. They left them operating well ahead of it.
That does not make the over-hiring a fiction. Between 2020 and 2022, Meta added 50,000 people, Alphabet added 70,000, and Amazon's corporate workforce grew to a peak of roughly 415,000, according to The Information. The hiring was built on a working hypothesis: that pandemic-era digital consumption rates were the new permanent baseline. Streaming, e-commerce penetration, cloud workloads, and ad spend all appeared to be entering a new structural regime. That hypothesis turned out to be wrong, and the hiring went into demand that never arrived.
Hiring as a driver of growth
I spent time on a related question that mattered to me: whether the hiring itself could have created the growth these companies were betting on. The intuitive case exists - you have to build before you can sell, software needs engineers before it ships - but the research holds a more disciplined answer. Gutiérrez et al. (2020), studying job postings across 14,000 public and private companies, found that increases in postings do correlate with future revenue and earnings growth. That finding has been cited as evidence that hiring leads growth. What it actually demonstrates is subtler: hiring decisions signal management's confidence in the demand pipeline. Boards act on private knowledge about contracts forming and products approaching launch by committing to headcount, before the financials confirm it. The data shows that the hiring leaks the information, it does not create the outcome.
The 2020 to 2022 data makes the distinction visible. Revenue per employee at both Meta and Alphabet declined during the period of most aggressive hiring, exactly the opposite of what a "headcount drives growth" model would predict. Apple, which grew headcount 24% while peers expanded 50 to 150%, tracked comparable revenue trajectories over the same period. And when Meta cut its workforce by 22%, revenue grew roughly 40% in the following two years. Hiring, it turns out, is evidence of conviction in a thesis.
The useful question it raises is not whether the people will generate the growth, but whether the thesis itself holds.
McKinsey's research on the dynamic captured the underlying logic plainly: shareholder value leads to jobs. The causation runs from stock performance to hiring conviction, not the other way around. Cascio's 37-year study of S&P 500 companies, published in the Academy of Management in 2021, reinforces this: firms that simply cut workers without structural change did not improve long-run returns. The cuts are a signal sent to the market, not a strategy in themselves.
The capital bet
The same five hyperscalers are now running a similar play, this time in capital expenditure rather than headcount. CreditSights projects top-five hyperscaler capex growing from approximately $256B in 2024 to $443B in 2025 and $602B in 2026. Capital intensity at Oracle has hit 57% of revenue. The underlying thesis is genuine, AI infrastructure investment is framed internally as the cost of relevance in the next platform cycle, and the risk asymmetry pushes everyone toward the same behavior, since underbuilding feels existential while overbuilding feels absorbable. IBM CEO Arvind Krishna stated publicly in December 2025 that "$8 trillion of capex means you need roughly $800 billion of profit just to pay for the interest." Moody's flagged rising overbuild risk in its $700B capex projection. Comparisons to the dot-com fiber buildout that left WorldCom and Global Crossing as cautionary tales have moved from analogy to mainstream concern.
Big Tech · Headcount vs Capex · 2019–2026
Two bets, two waves
Both series indexed to 2019 = 100. The people bet peaked at 162 in 2022. The capital bet is at 435 in 2025 — and 2026 committed guidance puts it near 900. Same companies. Same pressure. Same mechanism.
Click a company to add its individual headcount and capex index alongside the combined lines.
Sources: SEC 10-K and 10-Q filings; company earnings releases. Headcount: Apple, Microsoft (FY Jun), Alphabet, Amazon corporate (~22% of total headcount), Meta — each indexed to own 2019 baseline. Capex: PP&E purchases from cash flow statements. Amazon capex includes warehouses/fulfillment; 2020–22 spike is pandemic logistics, not AI. Amazon 2025 estimated from quarterly disclosures. 2026 capex = committed guidance: Amazon $200B + Alphabet $182.5B + Microsoft $190B + Meta $135B + Apple ~$15B ≈ $722B. Headcount 2026 not projected. Aggregate index base: 2019 = 100.
The repetition of the pattern points to something structural. The Magnificent Seven now represent 33 to 34% of the S&P 500 by market capitalization, up from 12.5% in 2016. In 2023 they delivered 62% of the index's total return and in 2024 roughly 60%. Across both years combined they returned 156% while the other 493 companies returned 25%. They trade at a forward P/E of 31x against 20x for the rest of the index, a 55% valuation premium that is contingent on continuing to deliver the next narrative quarter after quarter.
Being ahead of the curve requires showing visible commitment to the curve.
In 2021, that meant hiring at a pace that demonstrated belief in the digital-everything thesis. Today it means committing tens of billions in capex with promises of "notably larger" numbers next year, in Zuckerberg's recent phrasing.
The pendulum swing is not a failure of judgment. It is a feature of how this system is designed.
Companies valued primarily on narrative must invest visibly in each new narrative, overshoot when conviction runs ahead of fundamentals, and correct with sufficient visibility to reset market expectations for the next cycle.
That is not a flaw in how these companies are run, it is the operating logic of being simultaneously the largest companies in the world and the primary vehicles through which the market prices whatever comes next. The system demands the swing.
The more clarifying question, when the next correction arrives, is not whether they got the thesis wrong. It is what the correction tells you about what the market is now willing to believe.