☕ CoinBucha
Explainer · reviewed 17 July 2026 · observable data

Corporate Bitcoin treasuries in 2026: a record stack, and the split the market is now pricing.

By Jukka Blomberg — ex-CMO of two international crypto exchanges

Two things about corporate Bitcoin treasuries are true at the same time in mid-2026, and they get reported as if only one can be. The first: public companies collectively hold more Bitcoin than ever. The second: the equity premium that funded a great deal of that accumulation is compressing, and the companies are no longer moving as one bloc. This explainer holds both sides up next to each other, each figure dated and sourced, and takes no view on where any of it goes next.

Information, not financial advice. Nothing here is a recommendation to buy, sell, or hold any share, coin, or asset. “Record holdings” is not a bullish signal and “premium compression” is not a bearish one — they are observable, dated facts about balance sheets and market prices, cited so you can check them yourself.

Side one: the stack is at a record

As of 16 July 2026, the public-company treasuries dataset CoinBucha reads (CoinGecko) shows roughly 1,285,045 BTC held across 179 public companies — the highest aggregate on record. One name dominates: Strategy (MSTR) holds about 843,775 BTC, close to 4% of all Bitcoin, more than every other public company combined. The rest of the leaderboard, from the same dated dataset:

#CompanyBTC held
1Strategy MSTR843,775
2Twenty One Capital XXI43,514
3Metaplanet 3350.T43,000
4MARA Holdings MARA35,303
5Bitcoin Standard Treasury Co. CEPO30,021

Leaderboard as of 16 July 2026, CoinGecko public-company treasuries dataset (the same JSON the public-company treasuries hub renders from). Counts are point-in-time and can change; the live signal carries the current numbers.

And the buying did not stop in July 2026. Several holders added in the same fortnight, each purchase a dated public disclosure:

Side two: the split, and “the premium era is over”

The other half of the story is that these companies have stopped behaving as one trade. Coinpedia described it in July 2026 as the corporate-treasury strategy “splitting as corporate holders take different paths”: the same window that saw the buys above also saw sales. Riot Platforms reduced its holdings by 2,325 BTC in April 2026 to 15,680 BTC; Strategy sold about 3,588 BTC to meet obligations such as debt service and preferred-dividend payments even as it remained the largest holder; and MARA Holdings sold more than 18,000 BTC across March–May before adding roughly 1,000 back in June. Buyers and sellers, in the same cohort, in the same months.

The sharper framing comes from the financing side. In a 2026 report, DL News quoted John Fakhoury of Stacking Sats: “the premium era is over… we’re entering a phase where only disciplined structures and real business execution are going to survive.” That is attributed commentary, not a CoinBucha view — but it points at an observable thing. The engine that let treasury companies accumulate faster than they diluted was the equity premium: when a stock trades above the value of its coins, it can issue shares at that premium and buy more Bitcoin per share than it gives up. When the premium compresses toward or below 1×, that flywheel slows.

mNAV = equity market cap ÷ (BTC held × BTC price)

That compression is measurable. On CoinBucha’s own cohort methodology, in mid-July 2026 both Strategy and Metaplanet read at discounts — equity below the value of the coins — while a miner-plus-treasury name like MARA carried a premium that reflects its operating business as much as its coins. The point is not any single decimal, which moves daily and differs between trackers; it is that “a Bitcoin-treasury stock” is no longer one uniform bet. The live treasury-premium cohort carries the current per-name ratios and sources; the mNAV explainer covers how to read a premium or discount without over-reading it.

Why both can be true at once

A record aggregate stack and a compressing premium are not a contradiction. Holdings are a stock — the cumulative result of years of buying, which a few large sellers barely dent against Strategy’s 843,775-coin anchor. The premium is a flow condition — how the market prices the equity relative to those coins today, and therefore how cheaply new coins can be financed next. So the sector can sit at an all-time-high holding while the mechanism that grows it gets more expensive, which is exactly what “the split” and “the premium era is over” are pointing at from two angles. Reported in isolation, either half is misleading; together they describe a maturing, more selective market. That is a description, not a forecast.

How to read it without over-reading it

The honest use of these numbers is descriptive. Watch two things over weeks rather than days: the direction of the aggregate stack (is public-company Bitcoin still rising, and is it broadening beyond Strategy or concentrating further?), and the direction of individual premiums (are names drifting toward or away from 1×, and does that line up with issuance, debt, or a change in what a company is doing with its coins?). Neither tells you a discount will close or a premium will hold. And always check each figure’s as-of date and the tracker’s share-count and price basis before comparing two of them — the same company reads differently across methodologies on the same day.

See the live data

Both sides update on the free dashboard — aggregate holdings and the per-name premium/discount cohort, with current numbers and per-name sources, not the dated snapshot on this page.

Treasury-premium cohort →Public-company treasuries →mNAV, explained →More signal explainers →

Sources

General information about observable market conditions, not financial, investment, or trading advice, and not a recommendation about any share, coin, or asset. Holdings and premium figures are point-in-time, methodology-dependent, and drawn from public data on the dates noted — they can change and can differ between trackers. Verify current data before relying on it. How CoinBucha computes its cohort: methodology.