Compute commitments
Source records, code and data for When Compute Commitments Become Claims.
Download P01 source and data ZIP ↗Read P01 and its methods →Indiconomics / evidence desk
The Observatory
Work with the figures behind the research. Choose a scope, inspect a disclosure, or step through a scenario based on a filed tariff. Every view leads back to its source and limits.
03 data explorers02 replication packs
01 / Trade and production
India’s export value, examined.
Compare one of six overlapping export scopes across calendar years 2015–2022. Switch between current-price dollar values and a domestic-value share; the chart shows one scale at a time.
Selected series / 2015–2022
543.8current USD bnAll activities
Domestic value added · current USD bn
| Calendar year | Domestic value added (current USD bn) |
|---|---|
| 2015 | 332.1 |
| 2016 | 332.0 |
| 2017 | 371.7 |
| 2018 | 399.9 |
| 2019 | 404.0 |
| 2020 | 399.4 |
| 2021 | 481.5 |
| 2022 | 543.8 |
Scopes overlap and must not be added. A change between shares is in percentage points. TiVA is model-based; current-price changes alone do not establish real production capability, productivity or a policy effect. Domestic origin does not establish corporate ownership or where profits are retained.
02 / AI infrastructure
A useful-life revision, measured.
Each company reported a net-income effect after lengthening the estimated useful life of specified equipment. Inspect the disclosure against that company’s own reported net income.
Company periods and equipment populations differ. Select a company to see its own disclosure.
Year ended 30 June 2023
4.1%
Disclosed effect ÷ reported net income
The reported denominator includes the effect. The disclosed numerator is rounded.
0%100% of reported net income
- Disclosed increase in net income
- About $3bn
- Reported net income
- $72.361bn
- Equipment category
- Server and network equipment
- Estimated useful life
- Four years → six years
This is a disclosed historical accounting effect. It does not measure AI profit or a share of earnings growth. The numerator is rounded, so the displayed 4.1% ratio should not be treated as greater precision.
03 / AI infrastructure
Step through the tariff scenarios.
Choose one of 15 source capacity rows. These are assumed contract capacity scenarios for the post-ramp affiliated new-load capacity component, with an 85% ceiling.
Discrete source rows only. Selecting a row does not interpolate between capacities.
ASSUMED CONTRACT CAPACITY SCENARIOS / source row 08 of 15
Contract capacity100 MW
Capacity component82.5 MW
Component share82.5%
0%100%
The dashed marker shows the 85% ceiling. Post-ramp affiliated new-load capacity component only. The selected row is an assumed contract scenario, not measured customer load, actual use, total billing demand or a forecast. Displayed MW and percent figures are rounded from the source row.
Source: AEP Ohio Schedule DCT, sheet 223-4, effective 28 August 2026. This is the frozen tariff material used for P01, not a current tariff feed. The P01 pack records the source link; it does not redistribute the original filing.
Reproduce the work
Source and replication packs
These frozen files accompany the published working papers. Read the methods and source notes before reusing a result.
Export mix
Replication materials for When the export mix changes the story.
Download P02 replication ZIP ↗Read P02 and its methods →The Observatory / AI infrastructure
Follow the commitment.
Three places. Different instruments. Open a case to see what the documents establish.
Evidence snapshot8 September 2026
Research coverage03 cases
Markers identify research geographies, not individual facilities. Colour indicates the selected case.
Map: Natural Earth ↗Read the analysis
When Compute Commitments Become Claims
Electricity Demand, Customer Credits and Creditor Timing.
Open the working paperOhio / Power contracts
From requested power to a signed commitment
43.3%
Signed tariff capacity / capacity in study requests
5,642 MW signed13,022.7 MW requested
Commitments as of 12 February 2026
AEP Ohio reported that part of its studied data-centre demand had progressed to signed tariff commitments. That is a different milestone from requesting a connection.
Where the evidence stops
This measures a change in commitment status. It does not measure delivered electricity, operating servers or utilisation. The studied cohort is not all demand in Ohio.
These are different ratios answering different questions. They are not a ranking, a common loss measure, or a live market feed.
Return to the journalRead the research
The published catalogue
29 published reading pages across the programme. The cards follow the current published catalogue.
AI infrastructure
Who Financed the AI Site?
Customer-funded construction, fee credits and the value of timing.
Read article →AI infrastructure
The Server’s Second Life
Equipment-life revisions, reported earnings and the limits of extrapolation.
Read article →AI infrastructure
Two Public Stakes, Two Different Instruments
Norwegian financing and Andhra Pradesh incentives put different claims on public money.
Read article →AI infrastructure
When Compute Commitments Become Claims
Electricity demand, customer credits and creditor timing.
Read working paper →Trade and production
How Indian Is India’s Export Boom?
Separate the domestic value embodied in exports from its share of gross exports.
Read article →Industrial production
What Moved When C295 Assembly Moved to India
The same 56-aircraft order shows which production tasks crossed borders, and which dependencies did not
Read article →Trade and services
The Globalisation That Never Enters a Container
What remote delivery, foreign affiliates and embodied value each reveal about India
Read article →Strategic supply chains
The Distance Between a Mine and a Magnet
What company disclosures reveal about the stages of rare-earth supply that are operating, contracted, planned or still awaiting a customer
Read article →Industrial policy
A Subsidy per Job, or per Announcement?
A reconstruction across 14 sectors separates money paid from factory activity, and shows why neither establishes the price of creating a job.
Read article →Trade and climate
The Carbon Border’s Development Test
Cleaner production can lower a carbon charge. Contracts help determine who receives the saving, and whether proving it pays.
Read article →Trade and production
When the export mix changes the story
Domestic services in manufacturing exports across four selected economies, 2015–2022
Read working paper →Industrial policy and trade
Who Pays for Geo-economic Policy?
What company accounts, tariff studies and carbon-border contracts establish
Read working paper →Research
Fifty Per Cent of What?
India's semiconductor subsidy is quoted as a single number. Its controlling instruments use different cost bases and release support against eligible expenditure, with production obligations attached.
Read article →Research
A Floor Under What Price?
Three rare-earth arrangements quote $110/kg. Their terms say more than the shared number about risk, while inventory and diversified supply depend on how long a disruption lasts.
Read article →Research
Hapag-Lloyd’s Uneven Red Sea Year
In 2024, a carrier’s late recovery left annual operating profit flat while Suez’s receipts collapsed. The accounts reveal different outcomes, not a complete allocation of the disruption’s cost.
Read article →Research
The Cash Clock at Sea
What two 2022 balance sheets reveal about goods in transit, supplier credit and customer collection
Read article →Research
A Factory Bill With a Demand Floor
A March 2023 industrial invoice separates a charge for capacity from charges that depend on the hour of use.
Read article →Research
Who Pays for the Next Grid Connection?
A worked British connection example shows why the cost of building a link and the amount charged to its first applicant are different questions.
Read article →Research
Debt’s Second Bill
What India’s and Japan’s published bonds reveal about April 2026–March 2027
Read article →Research
The Different Clocks of a Rate Rise
The different clocks of a rate rise.
Read article →Research
What the Consumption Gap Measures
India’s 2022–23 and 2023–24 household surveys against one national-accounts vintage
Read article →Research
The Dollar Exposure Reserve Charts Miss
Two audited Indian company accounts show why the currency of a balance sheet, the date of a payment and the cash available to meet it must be read together.
Read article →Research
The Loan Had More Time. Its Valuation Had Already Fallen.
Eight observations of one private-credit holding show why a maturity extension, accrued income and an estimated recovery value must be read separately.
Read article →Research
A Digital Dollar Still Needs a Door to the Bank
USDC can move between wallets. Turning it into bank dollars, and receiving income from the assets behind it, depends on a different set of relationships.
Read article →Research
What the customer-service tenure decline can tell us
Four US January snapshots, 2020–2026
Read article →Research
Why the Same Industry Is Not the Same AI Investment
What two German information experiments can—and cannot—tell us about firms that hold back
Read article →Research
When AI Closes the Task Gap
What experiments reveal—and leave unresolved—about the value of a qualification.
Read article →Research
One rent, two pay packets
An April 2025 England budget scenario shows why the share of a pay difference left after housing depends on how many earners share the bill.
Read article →Research
When a Lower Remittance Fee Has a Higher Quoted Cost
A dated comparison of India-bound transfer quotes from the United States, United Kingdom and United Arab Emirates.
Read article →