🔍 Read the full analysis: What AI Performance Numbers Do—and Don’t—Mean on ThorstenMeyerAI.com
Get business pricing on monitors, keyboards and dev gear
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
A November 11, 2026, article from ThorstenMeyerAI.com uses GDP’s history to explain why a single economic or AI performance figure cannot capture all value. It points to omitted unpaid work and free digital services, then cites a Stanford working paper estimating $172 billion a year in consumer value from US AI chatbots. That estimate is not the same as revenue or a direct measure of model performance.
ThorstenMeyerAI.com published an analysis on November 11, 2026, arguing that familiar economic measures such as GDP—and headline figures about AI—can show only part of what people gain from technology. The article cites a Stanford working paper that estimates US consumers receive $172 billion a year in value from AI chatbots, while cautioning that consumer value, company revenue and model performance measure different things.
The article grounds its argument in the creation of national income accounting during the Great Depression. In a report sent to the US Senate on January 4, 1934, economist Simon Kuznets estimated that US income produced fell from $83.0 billion in 1929 to $39.4 billion in 1932, a 53% decline. The figures gave lawmakers a way to quantify the economic collapse, but Kuznets warned that national income could not by itself establish how well a nation’s people were doing.
Kuznets’s measure mainly counted activity whose results appeared in markets. It left out much household work, and did not show how income was distributed or how difficult it was to earn. The source article uses those limits to explain why GDP, a later measure of domestic production, should not be mistaken for a complete account of economic welfare or social value.
For a modern example, the article reports that unpaid household work in Germany was valued at €987 billion in 2013 when priced like paid work, equivalent to 39% of the value added counted in GDP. It also cites a 2017 experiment in which the median participant would have needed $17,530 to give up search engines for a year. That estimate reflects users’ stated willingness to pay, not money they actually spent. The article applies the same distinction to AI: a Stanford working paper’s $172 billion estimate concerns consumer value from US chatbots, not chatbot companies’ sales.
Infographic 01 · The sector clock, by output
China: how GDP changed shape
Share of GDP (value added at current prices) by sector, 1978 to 2025. In 2023 farming still held 22.8% of all jobs but produced 7.1% of GDP.
- Farming27.7%
- Industry47.7%
- Services24.6%
Reform starts
- Farming14.7%
- Industry45.5%
- Services39.8%
- Farming9.1%
- Industry45.4%
- Services45.5%
Services overtake industry
- Farming6.7%
- Industry35.6%
- Services57.7%
Communiqué 2025
Measured by output, not by jobs: the sector clocks of earlier episodes count workers. Jobs vs. output for farming, 2023: 22.8% of employment (year-end) vs. 7.1% of GDP.
Sources: NBS China, China Statistical Yearbook 2024, tables 3-2 and 4-2; NBS, Statistical Communiqué 2025 (28 Feb 2026).
Powered by Thorsten Meyer AI
Why AI Value Can Exceed Revenue
AI figures can answer different questions while sounding as if they measure the same thing. A benchmark score may compare a system’s results on a defined task. Revenue records payments to a provider. A consumer-surplus estimate attempts to represent how much users value a service beyond what they pay. These measures are not interchangeable, and none alone establishes whether a tool improves people’s welfare across everyday use.
The distinction matters as companies, investors and public agencies weigh AI adoption. A free or low-cost service may deliver substantial perceived value without generating comparable sales, just as GDP misses some unpaid activity and does not directly measure well-being. Conversely, a high score on a test does not by itself establish broad usefulness, reliability or economic benefit. Readers need to know the metric, population, task and comparison behind any performance claim before drawing conclusions from it.
The article’s examples also show why the choice of measure can shape policy and public debate. If decision-makers focus only on market transactions, unpaid care and services provided at little or no cost can disappear from the account. If they focus only on a consumer-value estimate, they may overlook who receives the benefit, who pays, and whether the estimate holds outside the study’s setting.
Products Worth Considering
EcoFlow Delta 3 Max Portable Power Station with 220W Solar Panel, 2048Wh
Samsung Galaxy Tab A11+ 8GB RAM, 256GB Storage, Gray
LG UltraGear 32GX850A-B 32-inch 4K UHD OLED Gaming Monitor
As an affiliate, we earn on qualifying purchases.
From Depression-Era Accounts to AI
The source article traces national accounting to a specific government need: in 1932, the US Senate asked the Commerce Department for estimates of national income amid the Depression. Kuznets led the work, and the January 1934 report offered a quantified view of the collapse. Its author also set out limits on what the total could say about welfare.
National accounts later developed to track broader production. The article says the United States added gross national product estimates in the 1940s, an international System of National Accounts appeared in 1953, and the US adopted GDP as its headline measure instead of GNP in 1991. GDP counts the market value of final goods and services produced within a country, commonly organized around consumption, investment, government spending and net exports.
The AI comparison is about measurement, not a claim that GDP and an AI benchmark are equivalent. The article says free digital services can provide benefits that are not reflected in their price, and argues that AI tools may follow a similar pattern. Its discussion of the Stanford estimate is an example of a valuation approach; the supplied source does not give the paper’s method, publication status beyond calling it a working paper, or detailed assumptions.
“A Stanford working paper estimates $172 billion a year in value for US consumers from AI chatbots.”
— ThorstenMeyerAI.com, summarizing its cited Stanford working paper
What the AI Estimate Does Not Show
The supplied source does not identify the Stanford working paper’s authors, title, release date, methods, sample or assumptions behind the $172 billion annual estimate. It is not possible from the provided material to assess how the researchers calculated consumer value, how uncertainty was handled, or how well the estimate applies to different users and chatbot products. The figure should be treated as a working-paper estimate, not an established accounting total.
The source also does not provide specific AI benchmark results or describe a newly released model, test or policy. Its subject is how to interpret performance and value numbers in general. A benchmark result depends on what was tested and under what conditions; without those details, no claim about a particular system’s ability or performance can be verified here.
Nor does the consumer-value estimate establish that users receive the same benefit over time, that benefits are distributed evenly, or that the services’ wider costs are included. The article’s comparison with GDP is a framing device. It does not show that the two measures use the same method or can be directly compared.
Check the Measure Behind Each Claim
The next step for readers evaluating AI claims is to look for the original paper or benchmark documentation. That means checking what the researchers measured, who or what was included, the time period, the comparison baseline and whether the result has been peer reviewed. For a dollar estimate, readers should also distinguish stated willingness to pay or estimated consumer surplus from actual spending and company revenue.
The supplied source announces no follow-up study, publication date or new benchmark release. Until more methodological details about the cited Stanford estimate are available, its $172 billion figure should be reported with its stated scope—annual value to US consumers from AI chatbots—and identified as a working-paper estimate. The broader question remains how to measure benefits that do not show up clearly in prices, revenues or test scores.
Key Questions
What is the main point of the article?
It argues that one number cannot describe every kind of value. GDP records market production but misses some unpaid work and does not directly measure welfare; AI performance scores, revenue and estimates of consumer value also answer different questions.
Does the $172 billion figure mean AI chatbot companies earn that much?
No. The source describes it as an estimate of annual value to US consumers from AI chatbots in a Stanford working paper. It does not describe the figure as company revenue.
What did Kuznets say GDP-like measures leave out?
His 1934 report said its estimates mainly covered activity visible in markets. It noted that unpaid household work was largely excluded and that income totals did not show how income was shared or how difficult it was to earn.
Does a strong AI benchmark prove a tool is useful?
Not on its own. A benchmark records performance on defined tasks under particular conditions. It does not automatically establish reliability in other settings, broad usefulness, user benefit or commercial success.
What is still unknown about the Stanford estimate?
The supplied material does not provide the paper’s authors, detailed methods, sample or assumptions. Those details are needed to judge how the $172 billion annual estimate was produced and how broadly it applies.
Source: ThorstenMeyerAI.com
