Bitcoin Market Structure

The comparison scores three ways of holding Bitcoin across six attributes that matter to large holders. Self-custody delivers high control and no counterparty risk, and pays for it everywhere else: usability is low, regulatory alignment is hard, scalability at size is poor, and inheritance is complex. Centralized custody inverts the trade. Usability is high, regulatory alignment is easy, and scale is good, but control is low and counterparty risk is high, which is the exposure that has repeatedly destroyed balances in this asset class. Multi-institution custody, which spreads key control across independent institutions so no single party can move funds alone, scores medium to high on control and minimal on counterparty risk while matching centralized custody on usability. It rates strong on regulatory alignment, excellent on scalability, and structured on inheritance. It is the only column without a weak entry in any row. The ratings are qualitative Early Riders analysis rather than measured values, so the useful output is the shape of the trade rather than any single rating. Most serious holders have already made this decision implicitly, usually by defaulting to whichever pole they encountered first. The table argues the choice is no longer binary. If a structure can hold counterparty risk near zero without forcing you to run your own operational security or leave heirs a puzzle, then treating security and usability as opposites becomes a choice rather than a constraint.

River estimates that individuals held 69.9% of total bitcoin supply in 2025. Every other category is a rounding error next to it. Lost coins account for 7.9%, funds and ETFs 5.9%, coins yet to be mined 5.8%, Satoshi's holdings 4.9%, businesses 3.9%, governments 1.5% and miscellaneous 0.9%. Add funds, ETFs, businesses and governments together and you get 11.3%, roughly one sixth of what individuals hold. This is an estimate rather than an audited register, since on-chain ownership can only be inferred, and the figures are River's work as published in Early Riders research in August 2025. The distribution matters because the story told about the last cycle was institutional adoption, and the ledger does not support it. Funds and ETFs, for all the flows, control less bitcoin than has simply been lost. That has consequences for anyone building or buying custody. The demand is not concentrated in a handful of large accounts that a single prime broker can serve. It sits with millions of individuals who need infrastructure built for their scale and their inheritance problems. Any service designed on the assumption that institutions are where the assets are is aimed at the smaller half of the market.

In 2020, the three largest categories of institutional bitcoin holders controlled 0.9% of circulating supply between them. By 2025 that figure was 14.5%. Exchange-traded funds and exchange-traded products account for the largest single share at 7.2%, which is striking given they held nothing at all in 2020 and 2021 and did not exist in US spot form until January 2024. Publicly listed companies hold 4.8%. Governments hold 2.5%, most of it acquired through seizure rather than purchase.

What changed is not just the quantity but the character of the holder. These are balance sheets and fund structures with reporting obligations, board oversight and, in the case of the funds, a creation and redemption process that converts retail savings flows directly into spot demand. They are slower to sell than the trading accounts that dominated the previous cycle, and in the case of government holdings, frequently prohibited from selling at all for long stretches.

Roughly one bitcoin in seven now sits in a structure that does not respond to a 30% drawdown the way a leveraged trader does. The float that actually clears the market is smaller than the circulating supply figure suggests, and it has been shrinking for five consecutive years.

US investment advisors oversee roughly $146 trillion. Fidelity, BlackRock, Morgan Stanley, Bank of America and JPMorgan have all published guidance recommending a bitcoin position of 1% to 5%. Actual advisor holdings sit at 0.008%, between 125 and 625 times below the range their own research departments recommend.

The third bar is not missing from the chart. It is drawn at 0.008% and is too small to see at a scale where 5% fits, which is a fair visual account of the situation.

The gap matters most for what it is not. This is not a case of institutions examining bitcoin and declining to hold it, which would be a considered judgment worth respecting. The firms publishing the recommendations are the same ones whose advisors have not acted on them. The distance between the two numbers measures operational friction rather than conviction: platform approvals, inclusion in model portfolios, compliance sign-off, custody arrangements, and the ordinary inertia of an advisory business that changes positions slowly.

That is the argument for treating market structure as the binding constraint rather than persuasion. Closing even part of this gap requires the plumbing to exist first, which is why custody arrangements and platform access set the pace far more than sentiment does. One percent of $146 trillion is $1.5 trillion.

The iShares Bitcoin Trust started trading on January 11, 2024 with no assets. Within months it was holding between $18 billion and $22 billion, a level it sat at through the middle of 2024. By December 2024 it had stepped up to $55 billion. Assets then fell back to about $40 billion in April 2025, since a fund's asset base moves with price as well as with flows, before climbing to roughly $95 billion at the start of October 2025. On October 29, 2025 it stood at $90 billion, 22 months after launch. No other exchange-traded fund has gathered assets on that schedule.

The number matters less as a milestone than as evidence about who is buying. A fund that size is not built by a handful of large tickets. It is built by continuous, mechanical demand arriving through ordinary brokerage accounts, retirement platforms and model portfolios that could not touch the asset at all before 2024. That demand does not negotiate on price and it does not stop when the spot market is quiet.

The access question is now settled. The question worth sitting with is what a permanent, price-insensitive bid does to a supply schedule that cannot expand to meet it.

About 6.56 million bitcoin are unrecoverable. The largest category by a wide margin is self-custody failure: 3.8 million coins, or 58% of the total, sitting in wallets whose keys no longer exist. Ponzi schemes and outright fraud account for 1.55 million. Exchange hacks, the category that generates almost all of the headlines, account for 1.2 million. DeFi exploits account for 7,500, which is a rounding error at this scale. Priced at $100,000 per coin, the full 6.56 million represents about $655 billion.

The ranking inverts the usual anxiety. The dominant risk to a long-term holder is not a sophisticated attacker or an exchange collapse. It is a hard drive in a drawer, a seed phrase written once and stored badly, or an owner who dies without leaving a recoverable path to the keys. Those failures are quiet, individual, and almost never reported, which is precisely why they compound to a larger number than every publicised theft combined.

The practical consequence is that custody design deserves as much attention as position sizing. Key redundancy, geographic separation of backups, and a documented inheritance process are the difference between owning bitcoin and having owned it.

Google Scholar mentions of http climbed from about 150,000 a year in 1980 to more than 2.6 million around 2012. Mentions of https crested near 2.2 million in 2014. Both then fell to roughly a quarter of their peaks by 2020, with http running under 500,000 and https near 600,000.

Nothing in that decline reflects failure. Early online commerce asked ordinary people to understand encryption, inspect certificates, and judge for themselves whether a payment page was safe, and most of them simply refused to enter card details at all. SSL solved the trust problem by making the cryptography invisible: a padlock icon and a protocol prefix replaced any need to understand the mechanism underneath. Once that worked, the question was settled and the research moved on. Infrastructure earns scrutiny while its outcome is contested, and loses it once it wins.

This is the pattern the parent article traces through personal computers, the iPhone, and SSL: a hobbyist era that demands real expertise, then institutional-grade infrastructure that hides the complexity, then mass adoption. Bitcoin custody is still in the first phase. Self-custody today asks of holders exactly what early e-commerce asked of shoppers, namely key management, operational security, and disaster recovery, with no forgiveness for a mistake. Multi-institution custody is the equivalent second-phase move, distributing trust across independent institutions while preserving on-chain verifiability and removing the single points of failure that centralized custody creates. The signal that it has worked will not be enthusiasm. It will be the moment the argument goes quiet.