Peter Wild

The Cuban claims register is public to the penny. Who owns the claims is not.

August 25, 2026

A friend floated a hypothetical this summer: what happens if Cuba opens up? There's a real answer buried in it. Helms-Burton makes resolving the certified US property claims a statutory condition of normalizing relations, so any settlement has to deal with a specific list of people who are owed a specific amount of money. That sounds like cheap convexity. Buy the claims at cents on the dollar, wait for a diplomatic event you don't control, get repriced.

I ran a quant fund for years, and the habit that stuck with me is that you don't get to have an opinion until you've built the dataset. So I spent a day building it.

What's in the file. The Foreign Claims Settlement Commission, a DOJ body, adjudicated the Cuban Claims Program between 1964 and 1972. The output is 5,951 certified claims across 7,048 claimant rows, published as a 109-page PDF on justice.gov. Total certified principal is $1.906 billion. The certificates carry 6% simple interest, so a rough constant-rate estimate puts it near $9.5 billion today, though every historical lump-sum settlement I found either haircut the interest or dropped it.

The parse was ordinary PDF archaeology with pdfplumber, plus the usual defects you find in a scanned government document: a claim number split across two lines on page 84, a lowercase cu-0811, an unhyphenated CU1968, one row printed twice. The parser logs each repair rather than silently fixing it.

The validation came from Richard Feinberg, who computed a total from the same FCSC list for Brookings in 2016 and published $1,906,479,883.33. My parse came to $1,906,479,883.94. Sixty-one cents apart across 7,048 rows, which is about as good a signal as you get that you read the document correctly.

The number everyone quotes for these claims is $1.851 billion. That's the FCSC's own 1972 program-report figure, and it is not what the FCSC's own published list adds up to. It's about 3% low. Feinberg's number and mine agree; the famous one doesn't.

The value is extremely concentrated. The top 10 claims are 53.8% of all certified value, the top 50 are 81%, the top 200 are 92.5%. The 835 claims worth $50,000 or more account for 98% of the money. Everything else is rounding.

The largest single claim is Cuban Electric Company, CU-2578, certified at $267,568,414. Cuban Electric was absorbed by American & Foreign Power, which went into Ebasco in 1967, which Boise Cascade bought in 1969, which renamed itself OfficeMax in 2004, which merged into Office Depot in 2013, which became ODP in 2020. The largest US claim against Cuba currently sits inside the company that sells you printer paper.

The obvious version of the trade is dead, and the data is what kills it. If you're thinking about buying claims, the tempting entry point is the small end, where sellers are unsophisticated and nobody's hired a lawyer. So I counted it. Roughly 4,200 securities claims under $1 million sum to $78 million, averaging about $18,500 each. They cluster by issuer, and the biggest cluster is Vertientes-Camaguey: 1,297 separate certified claims that together come to $9.45 million.

Sweep every Vertientes claim in existence and you have collected under $10 million of principal, after doing heir research, nationality diligence, and assignment paperwork on 1,297 estates. The per-claim cost exceeds the per-claim value at any recovery rate you'd care to assume. It's uneconomic on arithmetic alone, and you can only know that after you count.

Two Supreme Court decisions repriced all of this in 2026. In May, Havana Docks v. Royal Caribbean read "property which was confiscated" broadly and put the cruise lines back on the hook, nine figures apiece. In June, Exxon Mobil v. Corporación CIMEX held that Helms-Burton abrogates the sovereign-immunity defense for Cuban state entities, so Exxon's suit over its $71.6 million claim proceeds. Certified claims went from dormant paper to enforceable litigation assets in about five weeks.

None of that transfers to a buyer. Section 6082(a)(4)(B) excludes claims acquired for value after March 12, 1996 from bringing a Title III action at all. If you buy a claim today you get the settlement lottery ticket and none of the litigation right, which is the part that became valuable in 2026. The claims appreciated in a way that only reaches holders who have never sold.

Havana Docks is also the cleanest evidence that certified face value tells you almost nothing. Its claim was certified at $9,179,700 in 1971, nowhere near the top of the list. It's now potentially a nine-figure recovery, because the docks are infrastructure that cruise lines were actively using, and use is what Title III attaches to. What a claim is worth depends on whether somebody is standing on the property today, and the register has no column for that.

Nobody knows who owns these claims. The FCSC certified who was owed money in 1972, and its authority to track transfers of those claims expired along with the program. Sixty years of mergers, spinoffs, bankruptcies, dissolutions and estates later, no official record exists of who holds them today.

That gap has teeth. Continuous US nationality is a condition of a claim surviving, and a good number of the large claimants have since been bought by foreign parents. Chiquita went to Cutrale-Safra in Brazil, Lone Star Cement to Buzzi in Italy, International Harvester's line runs to Traton, Firestone to Bridgestone. Each of those raises a real question about whether the claim still exists, and no agency is tasked with answering it.

So the government maintains a register, precise to the cent, of what is owed, and no register at all of who is owed it. I mapped 49 of the top 200 claims to their present-day holders, covering about 79% of all certified value, with a confidence label on every row. The ones that resisted are the interesting ones: which ITT successor vehicle holds CU-2615, where the Amstar chain ended up after its stint under UK ownership, what happened to West Indies Sugar and Bangor Punta. That's entity resolution across six decades of corporate history, and it's the only part of this whose scarcity is real.

I didn't buy anything, and I'm not going to. Buying convexity on a diplomatic event with no date attached, in an instrument that gets worse the moment you own it, is a wish with a spreadsheet.

What I have instead is a clean dataset, a tail I can prove is noise, and an honest answer to a question I couldn't have reasoned my way to. It's the same thing I found with the trading experiment I wrote about in June. Building the apparatus is cheap now. Picking the question is the expensive part, and that's still entirely on you. This one took a day and came back no, which is a fine outcome for a day. The 61-cent match I'd have taken either way.

Sources. Everything here comes from public US government documents.

This is personal research, not investment or legal advice.

← Writing