Jake Bowser · Independent analysis
Governments are signing deals over minerals, electricity and computing capacity faster than anyone is analysing them. This takes one agreement at a time, clause by clause, and works out what it obliges, what it pays, and where the loss lands when the assumptions do not hold.
No. 01 · US–Ukraine Reconstruction Investment Fund
Ukraine assigned half the state’s take from every deposit it has not yet developed. On the base case, the first payment arrives in 2039.
The framework text is public. The partnership agreement that decides the economics is not — and the obligations run almost entirely one way. Three pieces, one open model, every assumption labelled.
2039
Modelled first production revenue, base case — fourteen years after the agreement took effect
0.85%
Ukraine’s entire subsoil rent line as a share of budget revenue, May 2026
2 of 4
Lithium deposits now outside Ukrainian control, both lost after the deal was struck
Signed 30 April 2025, effective 23 May 2025. One of the very few strategic-minerals agreements whose framework text is public — and one whose economics sit almost entirely in a partnership agreement that is not.
A clause-by-clause ledger of every obligation in the framework agreement — who bears it, which article creates it, and whether the language binds or merely expects. Fifteen obligations; the asymmetry is in the verbs.
The fund takes 50% of subsoil revenue from new licences. Establishing what that attaches to: the entire subsoil rent line is 0.85% of Ukraine's budget revenue, and the producing assets all sit outside the perimeter.
Modelling first revenue. Base case is 2039 — fourteen years after the agreement took effect — while US capital accrues automatically from military assistance the whole time.
Each analysis works only from documents that exist in public: contract text where it has been published, IMF and EITI reporting, development-bank disclosure, listed counterparty filings. Where something cannot be determined it is listed as unknown rather than estimated quietly. Every analysis contains the same six things, in the same order — the consistency is what turns a set of pieces into a methodology.
Obligation ledger. Every commitment, who bears it, which clause creates it, binding or aspirational. Government, project company and private sponsor kept separate.
Cash-flow model. Reproducible, with every input sourced and dated, and the assumptions table published beside the outputs.
Payment waterfall. Who gets paid in what order, and who absorbs the first loss.
Downside cases. Construction delay, price floor, output shortfall, currency. Where reserves run dry and where covenants trip — not one probability-weighted number.
Five terms worth renegotiating. Ranked by economic effect per unit of negotiating capital.
Unknowns register. What could not be determined, and which document would resolve it.
These are drafts published in the open. Assumptions are labelled as assumptions, models are in the repository, and corrections are welcome and will be made visibly rather than quietly. If a figure here is wrong, the fastest way to fix it is to tell me which input is wrong and why.
Jake Bowser. I am a mortgage lender by trade — VP of Mortgage Lending, working mostly in Non-QM, where the job is reading a term sheet, working out what it actually pays when the assumptions fail, and explaining that to someone who has to decide on it.
That is a smaller version of what this series does. The zeroes are different; the discipline is not. An offtake agreement and a loan term sheet are the same question asked at different scale: what does this actually pay, under what conditions, and who absorbs the loss.
I publish the models because analysis should be judged on whether it is right, not on whose letterhead it arrives on. If a number here is wrong, the input that produced it is in the repository and you can show me.
If you are working on an agreement of this shape and want the analysis done properly, that is the work I do.
Models and source data: development_curve.py, published alongside. Built as static HTML with no framework, no build step and no tracking.
© 2026 Jake Bowser. Analysis is offered for discussion and does not constitute investment, legal or tax advice.