PROTOCOL RESEARCH

Paid in your own paper

More than a hundred crypto projects wound down in the first half of 2026. Almost none of them were frauds, and almost none of them were killed by the hacks that got the headlines. The thing that killed them was sitting on the balance sheet the whole time.

The count depends on who is doing the counting — RootData had 99 by late July, other trackers passed 100 shortly after, with over half of them DeFi. What matters more than the number is the shape. Celsius, FTX and Terra vanished in days. These faded over months: user activity thinned, fee revenue shrank, the treasury compressed, and by the time a wind-down notice went out there was nothing left to organise a rescue around.

Most coverage filed this under bear market. That is true and not useful. There is a specific mechanism underneath, it was visible in advance, and a lot of protocols still have it.

The trade nobody called a trade

From roughly 2021 through 2024, small and mid-cap projects did not run on fee revenue. They ran on the appreciating value of their own token. Developers were paid in it. Liquidity was subsidised with it. Audits, legal work and marketing were funded by selling it.

Seen from the inside this feels like resourcefulness — you have an asset, you use it, and you avoid burning scarce cash. Seen from outside it is a position: the protocol is long its own token, with its entire operating runway as the collateral.

And it is a position with one unusual property. The value of the treasury and the ability of the business to raise more capital are driven by the same variable. When the token falls, the runway shortens, the team cannot sell into a thin book without accelerating the fall, and venture capital will not refinance an asset it has already written down. Every exit is the same exit.

Ordinary corporate treasuries are boring on purpose for exactly this reason. A company holding cash and treasury bills has a runway that does not care what its own equity did last quarter. A DAO holding its own token has a runway that is a leveraged bet on itself.

The number that should have been the story

Here is the figure I have not seen given the weight it deserves: around 42 percent of DAOs still hold more than half of their treasury in their own native token.

That is not a legacy problem being worked through. That is the current state, after a year in which this exact structure emptied a hundred projects. The lesson is available, priced, and demonstrated, and the majority of the industry's balance sheets have not moved.

If you hold a governance token, that ratio is a more informative risk metric than TVL. TVL tells you how much other people's money is currently parked. Treasury composition tells you how long the team can keep the lights on if the token halves.

Why the hacks were the symptom

The early framing of 2026 was a security crisis. Over $770 million was drained in the first months, with Kelp DAO at roughly $293 million and Drift at roughly $285 million accounting for most of it in April alone.

Both of those protocols are still operating. Meanwhile Step Finance shut down after a treasury hack initially reported near $27 million and later assessed closer to $40 million. Summer.fi wound down after an exploit of around $6 million. Carrot Finance closed after an $8 million loss left its reserves insolvent.

Look at those numbers next to each other. The protocols that survived lost forty times what the protocols that died lost. The exploit was never the deciding variable — the balance sheet behind it was. A well-capitalised protocol absorbs a nine-figure hack. A token-funded one does not survive a seven-figure one, because it had no reserves that were not correlated with its own share price.

The same logic explains why no white knights appeared. Rescue capital for a protocol with real revenue is an investment. Rescue capital for a protocol whose only asset is its own token is a donation to a falling knife, and venture funds had already marked those positions down. New crypto fund formation fell to multi-year lows, with roughly eight crypto-focused funds raising about $1.1 billion in the first quarter.

What was actually visible in advance

None of this required inside information. The leading indicators were public the whole time:

• Treasury composition. What share is the protocol's own token? Above half, the runway is a leveraged position.
• Revenue against cost. Fee revenue is published for most protocols. If it never covered costs, the gap was being paid by token sales, and token sales have a floor.
• Whether yield exceeded what the revenue model could support. Where it did, the difference came from emissions or treasury draws. Both are finite, and both were being reported as growth.
• Developer activity. Stale repositories preceded wind-downs with depressing reliability.
• What happened when incentives tapered. Everclear is the textbook case: around $500 million in monthly volume and no path to self-sustaining revenue once the incentives stopped. Volume that leaves the moment you stop paying for it was never revenue; it was a rebate with a marketing budget.

The uncomfortable generalisation

Token-funded operations are not a mid-cap problem. They are the default across most of the industry, and the large protocols differ mainly in having more of everything — deeper secondary liquidity, longer runway, more patient holders. That is a difference of degree, not of structure.

What 2026 demonstrated is where the degree stops being enough. Liquidity for a top-ten token does not evaporate the way it does for a mid-cap, so the same balance sheet buys years instead of months. The mechanism is identical. The clock is just slower.

Which leads to the question worth asking of any protocol you are exposed to, including the ones that look untouchable: if the token fell 80 percent and stayed there for eighteen months, what would this team be able to pay for? For a business with fee revenue and a diversified treasury, the answer is most things. For everyone else, the answer is the one a hundred projects gave this year.

The honest summary

2026 was not a fraud year and it was not primarily a hacking year. It was the year the industry's standard financing model met the conditions it was never designed for, and the bill arrived at a hundred addresses more or less at once.

The fix is not complicated and it is not new — hold a treasury in assets uncorrelated with your own success, and charge enough that revenue covers cost. Both have been obvious for years. Both were ignored while the token went up.

Roughly two out of five DAOs are still doing it the old way. The next time secondary liquidity thins, that number is the list.

Analysis, not investment advice. Figures reflect public reporting through mid-2026 and shutdown counts vary by tracker and by what each one counts as a closure — verify before relying on them.