What is the
Intelligence Hours Index?

Intelligence Hours · By Kwangseob Ahn, INLEVEL9

The Intelligence Hours Index (IHI) measures AI subscription affordability by country: the gross minimum-wage work required to pay for a specified monthly AI plan. It is a labor-time price index, not a benchmark of AI model capability.

Buying access to dollar-priced AI gives us another international price to watch: the price of intelligence. The supplier names a dollar price; workers meet it with local earnings. The question is how much of their working time that price requires. Read the argument and its sources.

The US July 2025 export order promotes American AI technologies and standards worldwide. That provides policy context for the author's question, rather than evidence of reserve-currency effects: when intelligence has a dollar price, how much can a country's labor buy?

I propose the Intelligence Hours Index (IHI) as a new Big Mac Index for the AI age. It translates the dollar price of a specified monthly AI subscription into gross statutory minimum-wage working time. The supplier names a price in dollars; workers pay for it with their time. A lower IH means that less time is needed. Improving that purchasing power asks about wages, exchange rates and how the gains from productivity reach workers.

1. Definition and data

For country i in year t:

Here P is the monthly subscription price in US dollars, and w is the selected gross minimum-pay reference per hour, converted to US dollars. Written in local currency, the same calculation is:

The research uses a pre-tax USD 20 common-price benchmark for 2023–2025. The website’s fixed three-plan basket uses the median US monthly list price of ChatGPT Plus, Claude Pro and Google AI Pro: USD 20, USD 20 and USD 19.99 at the September 28, 2026 review. Its median is USD 20, before taxes. This benchmark makes the price comparable; actual local checkout prices, taxes and plan availability can differ.

Public v5 audits dated statutory wages and their conversion units, adds consumer-tax scenarios, and corrects the local-price comparison for tax treatment. The historical headline uses selected adult basic-pay references, with mandatory extra payments studied separately. Canada's and Japan's historical values retain employment-weighted regional benchmarks; Belgium's guaranteed average income has a distinct remuneration scope. Consumer purchasing power needs net take-home pay, while an employer's labor-cost comparison needs gross pay plus employer contributions. Those denominators answer different decisions and require matched evidence.

The historical study covers 37 countries, retaining frozen OECD-published observations alongside dated primary-source corrections over 2023–2025, with market-rate IH available for 36 in 2025. At USD 20, the 2025 range runs from 1.1 hours in Luxembourg to 16.2 in Brazil, a 14.18-fold pre-tax spread. Luxembourg's 17% VAT scenario gives 1.34 hours. Brazil's service-tax provisions depend on the transaction route and municipality, so a single all-in checkout rate is not assumed. Its 2025 ordinary minimum-wage payroll model gives R$1,404.15 net from R$1,518 gross. Applying that factor to the comparative hourly benchmark gives 17.51 hours before consumer tax. Luxembourg's selected payroll factor applied to the comparative hourly benchmark gives 1.30 hours before VAT; their selected-profile pre-tax ratio is 13.51. These are conditional factor applications, not exact matched payroll observations. The 35-country standard-rate subset ranges from 1.34 to 15.50 hours after its standard-rate adjustment and excludes Brazil. The website has a broader 50-country catalog and a daily reference exchange-rate update schedule. Each result shows its actual input dates and wage jurisdiction; countries with missing or non-comparable wage data remain explicitly marked.

The same ratio applies to higher prices. At USD 200 a month, the following wage references give very different labor-time prices:

Historical and scenario figures on this page come from the companion public v5 paper and its revision analyses and source inputs. The original Zenodo 1.0.0 deposit preserves the earlier baseline. Figures retain their research reference dates; the current calculator shows the latest published inputs. Calculation rules and source details are in Methodology & Sources.

2. A Big Mac Index for the AI age

The Big Mac Index makes an international economic relationship legible through a familiar purchase. That is the ambition here, too: use an AI subscription to ask what a country's labor can buy in a global market. It is an invitation to compare and to ask how to improve, with working time as the common unit.

The calculations answer different questions. The Economist compares local burger prices and exchange rates to illustrate purchasing-power parity. A burger's price includes local wages, rent and other inputs. IH holds the subscription's dollar price common and measures the local work needed to buy it. IH describes local labor facing a common global price. Calling it a new Big Mac Index describes its public role; IH does not estimate whether a currency is overvalued or undervalued. With a common price, ranks are exactly the inverse of dollar-converted hourly wage ranks. Read The Economist's calculation and data.

3. Access prices and task costs

Task-based automation theory asks whether a machine can produce an equivalent output at a lower cost than human labor. That question motivates interest in AI access, but a monthly subscription fee does not measure the cost of completing a particular task to a given standard.

IH measures the wage-floor price of access, not an automation threshold. A task-level cost comparison would also need the affected occupation’s wage, relative productivity, output quality, usage limits, supervision and integration costs. A statutory minimum wage is not every employer’s actual labor cost.

The hours reported by IH are earnings needed to pay the subscription fee. They are not hours saved by using AI. The index does not measure intelligence, model performance, productivity or human worth.

4. How can a country improve?

A ranking invites the question: how do we move up? For IH, the practical goal is to reduce the time needed to buy access. Raising the statutory minimum wage is one route, not the only route. An unchanged wage can buy more dollar-priced intelligence when its currency appreciates. A wage increase can buy less when depreciation more than offsets it.

Productivity opens a further question: do the gains reach workers' international purchasing power? Better skills, technology or production can matter through their effects on the applicable wage floor or exchange rate. Productivity is not a direct input to IH. If neither the wage floor nor the exchange rate changes, a productivity gain alone leaves the headline hours unchanged. The ILO's wage-setting guide discusses productivity and sharing the gains of progress. Their transmission matters here. Raising the legal wage floor lowers the benchmark mechanically; for the gain to reach a worker's actual purchasing power, that wage must apply and be paid.

This is a way to read outcomes, rather than a prescription to target a stronger currency. Exchange rates respond to many forces, currency strength is not a complete measure of national competitiveness, and higher productivity does not automatically cause appreciation. Domestic AI supply or lower regional prices may also improve actual access, but comparing them requires separate evidence on price and capability.

With P held constant, the change in IH decomposes exactly. Write the exchange rate as local currency per US dollar, and the minimum wage in local currency:

This identity accounts for IH movements through exchange rates and wage references. A wage reference may also change with working-time rules, remuneration scope or the composition of a weighted regional benchmark. The terms do not isolate causal effects. If the dollar price changes, its log change must also enter the equation.

Between 2023 and 2025, IH fell in 33 of the 36 countries with comparable observations as wage growth and exchange-rate movements combined to reduce the labor-time price. Korea illustrates the opposite outcome. Its minimum wage rose by 4.2 log points in total, while the won depreciated by 8.6 log points. The currency movement offset roughly twice the wage-policy gain.

For Korea’s 2026 hourly minimum of ₩10,320, an exchange rate of about ₩1,463.6 per US dollar would return IH to its 2025 annual-average level, erasing the affordability gain from the 2026 wage increase. This is a calculated threshold against that historical baseline, not a forecast. Where wages are set once a year and exchange rates move daily, the effective AI purchasing power of a wage is determined jointly by both.

Lower hours and a better rank are different achievements. A common dollar-price reduction lowers every country's hours proportionally without changing their order in an unchanged sample. Another country's improvement can change your rank even if your own hours stay the same. Read the hours and the wage–currency movements alongside the ranking.

5. Does regional pricing close the gap?

Only partly in the paper’s dated pricing pilot. ChatGPT Go launched in India at ₹399 a month as a separate, lower-cost tier. A cheaper tier with different model access, features and usage limits is a different offer; it does not establish a discount on the same plan. OpenAI’s Go plan documentation describes those plan distinctions.

Using 2025 average exchange rates and a USD 8 US reference price, the pilot puts the observed Go prices in India, Indonesia and Nigeria at about 57% of the US level. Korea’s April 2026 observation, ₩13,000 including VAT, is about 114%, or 104% net of its reported 10% VAT. Plus prices must receive the same tax adjustment before comparison with the pre-tax USD 20 US reference. Removing tax gives Plus ratios of 0.93 in Korea and, conditionally on tax-inclusive lists, 0.97 in India and 0.95 in Indonesia. The Indian and Indonesian tax inclusion is unverified. The revised reading is approximate pre-tax parity under these scenarios, alongside a much larger Go listed-price discount.

These are historical list-price comparisons, drawn from reports of provider statements. Tax treatment is not consistently documented, and local checkouts were not independently verified. The US Go reference price dates from January 2026; the pilot does not establish that the 2025 local prices were still in force on that date. The pilot supports a narrower conclusion: lower entry prices need to be compared within the same plan before they can be called cheaper access to the same capability.

6. From affordability to use

In the USD 200 scenario above, the subscription costs 27.6 hours of minimum-wage earnings in Korea, 12.7 in Germany and 97.7 in Mexico. These are differences in the wage-floor access burden. They do not identify who subscribes, how much useful work a tool delivers or whether an employer replaces labor.

India illustrates why the wage jurisdiction and date matter. The paper withholds an unverified national wage range. Delhi's ₹710 daily unskilled floor, effective from April 2025, remains a dated regional illustration whose exact continuation at the January 2026 cutoff is not independently established. China's separate regional part-time examples also do not establish a harmonized national range. Japan's snapshot uses prefectural floors already in force on 1 January and withholds an unverified date-matched national point. Read the snapshot's units, dates and scope limits.

The public paper's supplemental check asks whether IH adds information about country-level Claude usage after accounting for national income. A common-price IH cross-section already mirrors dollar minimum wages and is strongly associated with income, so this is a limited incremental-information check. The outcome includes free usage and is not a paid-subscription adoption rate or a measure of uptake among minimum-wage workers. This result does not prove that affordability has no effect. Read the external-criterion check.

This check belongs to the updated public paper, which is distinct from the journal-submission manuscript. Its V5 supplement includes revision analyses and the AEI check. The Zenodo 1.2.0 data-and-code deposit includes the public v5 country-source correction; the earlier Zenodo 1.1.0 revision preserves the 2 October tax and statistical edition, and the original Zenodo 1.0.0 deposit preserves the earlier baseline.

7. Your turn

What would let your country's workers buy the same access with less of their time? What would transmit productivity gains to their purchasing power? A useful index should invite these questions and scrutiny of its inputs. Four limits matter especially:

• Minimum pay as a reference. It does not represent typical pay. In the paper's 32-country κ-rescaling scenario, rank correlations are ρ = 0.949 and τ = 0.823. The median absolute rank change is two places and the upper quartile three; the United States moves ten places and the United Kingdom five in the opposite direction. The same-year source ratios are not fully matched to the revised hourly wage inputs, so these results are a scenario rather than an observed median-wage ranking. Chile has no same-year 2025 ratio and is excluded.
• Single price point. IH tracks a consumer subscription benchmark. Enterprise contracts and usage-based API bills require separate price measures. Free AI services also exist.
• No quality adjustment. Plans with the same price can differ in capability, and capability changes fast. A fixed basket keeps the price comparison legible without asserting that its products are equivalent. A provider's flagship label is not an independent capability test.
• Coverage and enforcement. Statutory minima are not equally binding across countries. Regional schedules, covered workers and enforcement affect what the wage means.

Economists have long measured wages in baskets of goods, such as how many loaves of bread an hour of work buys. IH applies that labor-time perspective to the price of a specified AI subscription. It is easy to compute, can be updated daily, and separates nominal-wage movements from currency movements under a fixed dollar price.

We publish the price, wage, exchange rate, jurisdiction, source and date behind each website result. The basket is versioned so product changes do not silently rewrite earlier measures. Corrections, better sources and alternative interpretations are welcome.

Created by Kwangseob Ahn, at INLEVEL9.

Kwangseob Ahn is an adjunct professor in the School of Business at Sejong University in the Republic of Korea, where he teaches business data management, business analytics and statistical approaches to data-informed decision making. He has more than 15 years of experience in the IT industry and is the founder and principal consultant of INLEVEL9, a boutique consulting firm.

INLEVEL9 profile · LinkedIn · ORCID · Read the research paper

Kwangseob Ahn (안광섭), also known as Oswarld, created Intelligence Hours at INLEVEL9. These names refer to the same person. It began with a simple observation: a common AI subscription benchmark requires very different amounts of work across countries because the wages used to pay for it differ.

The index turns that gap into a number anyone can read: the hours of work needed to afford a month of paid AI. The aim is to give readers, journalists and researchers a lasting reference for discussing access to AI, as familiar and easy to compare as other international price indexes.

This is an independent project. It is not affiliated with the OECD or the AI providers listed here. Each published observation identifies its sources, reference dates and wage jurisdiction.

Questions, contributions and citations.

For questions about the index, corrections, additional official wage or price sources, research collaboration, or citation requests, email about@intelligence-hours.org. When suggesting a data update, include the official source URL, jurisdiction, covered population and effective date.

When citing a result, identify Intelligence Hours, the country and wage jurisdiction, the observation date, and the basket or individual plan. Link to the result and its sources so readers can reproduce the comparison. Contact us for publication enquiries or help interpreting a particular result.