Invest1 publisher3 min readPublished
Amazon earned $6.3m of profit for each sampled worker it has on food stamps
The GAO counted 12,346 Amazon workers on SNAP across eleven states, close to triple its 2020 finding, over a stretch in which the company's annual profit went from $11.6bn to $77.7bn. Labor's share of output is 52.8%.
The Investor · Invest desk
What happened
- Labor's share of US economic output has fallen to 52.8%, the smallest reading since the Bureau of Labor Statistics began tracking the series in 1947, according to Fortune.
- A GAO report counted 12,346 Amazon workers on SNAP and 11,338 on Medicaid across the eleven states it sampled, close to triple the number of Amazon employees on federal assistance it found in 2020.
- Over the same period Amazon's annual profit rose from $11.6bn to $77.7bn, and 2025 revenue rose 12% to a record $717bn.
- Walmart and FedEx recorded similar increases in workers taking federal assistance, as did rideshare and delivery companies.
Compiled by The InvestorSomething wrong?How this is made
Why it matters
- cost With SNAP eligibility set at roughly 130% of the poverty line, the taxpayer is covering the shortfall for people already working full time, and the employer's payroll expense is unchanged by that transfer.
- exposure GAO's tallies are broken out by employer, so Amazon, Walmart and FedEx can each be named individually in a benefits-cost argument in a way an aggregate labor-share statistic never allows.
- constraint If Stansbury is right that fissured employment is the main driver, naming one large employer reaches only its direct payroll and leaves the contracted workforce doing the same work outside the count.
- contradiction Anyone using labor share as a wage-budget input has to choose between two series 9.1 points apart, and the choice decides whether the picture looks like a 1947 low or a slow four-point drift.
Divide Amazon's $77.7bn of annual profit by the 12,346 of its workers the GAO found on SNAP in eleven states and the answer is about $6.3m per worker [6][4][1]. Fortune reported the count for those eleven sampled states, so the national number is larger [4]. The SNAP and Medicaid tallies add to 23,684, and because one worker can be enrolled in both, that total is enrollments and not necessarily 23,684 people [4][2]. A near-tripling since 2020 implies a base of roughly 7,900 in the same sample [5][5].
The company side moved faster than the count. Profit went from $11.6bn to $77.7bn, a factor of 6.7 [6][3]. Revenue rose 12% to a record $717bn, and $77.7bn on $717bn is a 10.8% net margin [7][4].
Two measures of labor's take appear in the same account and they do not agree. The BLS series puts labor at 52.8% of output [1]; KPMG's February work puts employee compensation at 61.9% of GDP, down from 66.6% in 1982 [15]. Those sit 9.1 points apart [6]. The market comparison has a related problem: a 600% gain in the S&P 500 since the start of the century against 12.5% growth in wages is 48 to 1 [2][3][9], but only the wage figure is adjusted for inflation [3].
KPMG's own pair does not net out. Corporate profits gained 7.85 points of GDP since 1982 while compensation lost 4.7, so about three points came from elsewhere in national income [15][7]. Diane Swonk, KPMG's chief economist, said of the chart, "Inequality fuels social and economic instability" [16].
Kathryn Larin, who directs education, workforce and income security work at GAO, told Fortune that today's safety-net recipients are overwhelmingly in the workforce and mostly full time [12]. "I mean, these are families that are really barely able to make ends meet, and yet they are working, and they are working a lot," she said [14].
Amazon disputes the reading. Spokesperson Rachael Lighty told Fortune the conclusion drawn from the GAO report is "wrong" and that looking at raw numbers instead of percentages is misleading [8], and said that "74% of our regular full-time employees are enrolled in an Amazon health insurance plan, well above the 65% private-sector take-up rate for full-time workers" [9]. Fortune's account does not include Amazon's headcount in the eleven sampled states, so the percentage the company says is the right measure cannot be computed from the published figures. Amazon did not respond to Fortune's request for comment [19].
Union coverage fell from 20.1% of US workers in 1983 to 10.0% in 2025, roughly a halving [17][8], and Anna Stansbury of MIT Sloan attributes the labor-share decline more to the breakdown of direct employment than to that [17][18]. An employer-level count is the part of this record that can be written into a bill, because it names a payroll. Nothing in Fortune's account points to a sponsor, a hearing or a measure, so the legislative step is a forecast. The counts could also fall for a reason unrelated to pay, if eligibility tightens; Amazon could win the measurement argument, after which the series gets reported as a share of headcount and a large employer looks unremarkable. In my view the first is likeliest, and what would settle it is a later GAO update showing counts falling while the labor share stays near 52.8%.
What to watch
- Whether GAO extends the sample beyond eleven states, which would give a national employer-level count for Amazon, Walmart and FedEx.
- Whether Amazon publishes the sampled-state headcount that would let anyone compute the percentages Lighty says are the fair measure.
- Whether SNAP eligibility rules change, since tighter rules would cut the counts without any employer raising pay.