About
One desk. One principal.
Halvren Capital is a one-principal research desk based in Vancouver. It covers twenty operators in Canadian and U.S. energy, materials, and infrastructure, deeply, with no managed product and no outside capital. The difference, where there is one, is that the analyst spent the prior decade operating businesses and underwriting them — not writing about them from the outside.
The principal is Amirali Karimi. Persian-Canadian, raised in Vancouver. My family has built businesses in British Columbia for more than thirty years — manufacturing first, then real estate. The last decade of my own work has split two ways. Operating: Karimi Developments on the real estate side, Tablo on the software side (a venture sold to Digikala, Iran's largest e-commerce platform, in 2023), and Boost Commerce Group, a Canadian digital holding. Investing: on principal capital, through every Canadian and U.S. cycle since the late twenty-teens. Economics at Simon Fraser University, in progress. BCIT Marketing Management diploma, earned earlier. IFC (Investment Funds in Canada) passed April 2026, CIRO exams in progress, CFA candidate*. The point of listing this is not that any single credential matters. The point is that the years it would have taken to learn the framework from the outside are already spent.
What “AI-augmented” means at Halvren is mechanical. Models read filings, transcripts, and pricing across the universe at a cadence one analyst could not match by hand. They surface what changed, what was said, who bought, who sold. The principal reviews every conviction, signs every position, and writes every public letter. There is no proprietary model, no internal LLM, no algorithmic alpha. The desk uses widely available tools at the volume the work requires. The architecture lives on the Process page.
The sectors are chosen because they are where boring, long-cycle businesses still compound — Canadian heavy oil, gas pipelines, regulated utilities, copper, lumber, fertilizer, rail — and because they sit far enough from the institutional consensus that careful reading still pays. The Canadian weighting is structural: it is easier to know a Canadian operator from the inside than to guess at a Texan or an Australian one. The U.S. names are added selectively, where the operator quality reads cleanly through the same lens.
How to read Halvren: the public research is free, including the Halvren Checklist, the 20-operator coverage universe, the research archive, and the Halvren Notes. The quarterly Halvren letter goes through Substack, where the email list lives; the site itself collects nothing. Halvren manages proprietary capital and is not currently accepting outside investors. The private letters page explains where that posture sits today and where it does not.
Most of what the desk studies never becomes a position. Most of what becomes a position is held for a long time. The writing reflects that pace.
What this firm is and isn't.
Halvren manages proprietary capital and publishes research. The capital is the principal's own. The research is informed by the same work that informs the capital, and is published openly so that anyone can read it, argue with it, or ignore it.
Halvren is not a registered investment adviser, broker-dealer, or portfolio manager. The desk does not manage outside capital, give personal advice, or sell a managed product. Nothing on this site is a recommendation to buy or sell any security. The full legal posture lives on the Terms and Privacy pages.
* Per the CFA Institute Standard of Professional Conduct VII(B), “CFA candidate” is used only while registered for the next sitting of a CFA exam. The references on this site reflect an active registration; if a sitting is missed without re-enrollment, the references will read “studying toward the CFA” and this footnote will record the change.
AI & indexing policy.
The thesis of Halvren is that machines read at scale. The site is built to embody that. Public research is free for individual readers and for non-commercial educational use, including ingestion by language models. AI training and machine retrieval are permitted with attribution to halvrencapital.com. A concatenated long-form bundle for offline ingestion lives at /llms-full.txt; an llmstxt.org-style index of the public surface lives at /llms.txt. The full /robots.txt spells out which crawlers we welcome (the major search-engine and AI-assistant crawlers) and which we do not (commercial scrapers and SEO link-graph harvesters that do not return value to readers).
Two things we do not do, on principle: there is no analytics, no tracking pixels, and no third-party JavaScript on this site, and we do not collect emails on the page (subscriptions go to Substack directly). Privacy is part of the brand, not an afterthought. Commercial republication, paid syndication, or automated downstream resale of Halvren research requires written permission — write if that is what you have in mind.