The MiCA blind spot: pricing a stablecoin major EU exchanges no longer list
Not MiCA-compliant, still in your NAV: where does your DAI price come from?
Date
Author
Gautier Humbert, Research Analyst
Every month-end, a crypto fund has to answer a simple question: what was each asset worth at the cut-off? Most of the debate focuses on the method: last price, TWAP, VWAP, mid-price. A more basic question comes first, and is often skipped: which market does the price come from?
The September 2026 close gives a clean illustration. Not on a volatile altcoin, but on DAI, a stablecoin designed to trade at $1.
DAI is not a niche token. With a market capitalization of about $4.5 billion, it ranks among the top 25 crypto assets (#22 on CoinMarketCap at the time of writing). Yet its daily trading volume across all venues is only around $17 million: a large asset with surprisingly little trading behind its price.
The setup
Asset: DAI/USD
Measurement time: 30 September 2026, 14:00 UTC, a common month-end cut-off (16:00 Luxembourg time)
Markets: the two direct DAI/USD markets in our coverage, Kraken and Gemini. Both are regulated, well-known exchanges.
On paper, the two closing prices look close:
Kraken: $0.99994
Gemini: $0.99771
A 0.22% gap. Nothing alarming for a stablecoin. But the trades behind these two numbers tell a very different story.
Context: DAI is leaving centralized exchanges
This close did not happen in a vacuum. DAI is in the middle of a transition:
Sky, the protocol formerly known as MakerDAO, has launched USDS as DAI's successor. The two convert 1:1, and DAI still exists on-chain.
Major platforms have started converting their clients' DAI into USDS: Crypto.com in April 2026, Coinbase in May 2026.
In Europe, DAI is not MiCA-compliant. Binance halted DAI spot trading for EEA users in March 2025, and Kraken did the same for its EEA clients in May 2025, converting remaining balances into USDC. Gemini had already stopped DAI trading and deposits for EU and EEA customers in February 2025.
The result: DAI liquidity on centralized exchanges is thinning out. Order books get shallower, single orders move the price, and some markets go quiet for hours or delisted. Keep this in mind for what follows.
Two markets, two very different pictures
Two markets DAI/USD are still available: Kraken and Gemini.
Over the previous 30 days (31 August to 29 September):
Kraken traded 4,035,003 DAI
Gemini traded 672,746 DAI, six times less
On 30 September, between 04:00 and 14:00 UTC:
Kraken printed 83 trades, spread across the day. The last one came at 13:56:55, three minutes before the cut-off: 556 DAI at $0.99994.
Gemini printed 31 trades. Thirty of them happened in the same 32 minutes, early in the morning.
What happened on Gemini that morning
At 04:31:42 UTC, a sell order swept Gemini's DAI/USD order book, pushing the price from $0.991 down to $0.985 in a single second.
Over the next 32 minutes, until 05:03, buyers took about 675,000 DAI, almost all of it at $0.98, 2% below the peg. That is more DAI than Gemini had traded in the previous 30 days combined.
Then the market went quiet.
At 08:01:27, one more trade printed: 1.9 DAI, about $1.90, at $0.99771. On its own, that trade moved Gemini's last price up by 1.8%, from $0.98 back near the peg.
After that: nothing. No trade for six hours, until the 14:00 cut-off.
So Gemini's "closing price" was set by a $1.90 trade from the morning. Had that trade not happened, Gemini's closing price would have been $0.98, 2% below Kraken. The 0.22% gap we saw at first was not a sign of agreement between markets. It was the result of one tiny trade.
Why the choice of market decides the price
Fair value frameworks such as IFRS 13 do not start with a formula. They start with a market. The price should come from the principal market for the asset: the market with the greatest volume and level of activity, among those the fund can access. Only when no principal market exists does the framework fall back to the most advantageous market.
This matters because crypto assets trade on many venues at once, with no consolidated tape. The same token can have a dozen "last prices" at the same second. Picking one is not a technical detail. It is a decision that directly sets the number in the NAV.
Applied to DAI on 30 September:
Volume: Kraken traded six times more DAI than Gemini over the previous 30 days.
Level of activity: Kraken traded throughout the day, up to three minutes before the cut-off. Gemini's activity came in one burst, then stopped for six hours.
Conclusion: on volume and activity, Kraken would be the principal market. Gemini is not, even though its 30-day volume looks respectable on paper.
Volume over a period says nothing about how that volume is distributed. A market can look liquid on a monthly report and be empty at the measurement time.
The principal market your fund cannot access
There is a second trap, and it matters most for European funds. The principal market is assessed from the fund's point of view: it must be a market the fund can actually access at the measurement date.
Both markets in this case are closed to European clients. Gemini stopped DAI trading and deposits for EU and EEA customers in February 2025, and Kraken did the same in May 2025. A Luxembourg fund trading through the European entities of these exchanges therefore cannot use either of the two markets that produced a DAI/USD price on 30 September. The sources it can still reach are of a different kind: on-chain liquidity pools, the 1:1 conversion into USDS through the Sky protocol, or a non-European exchange entity if its contracts allow it.
In that case, "pick the market with the most volume" is not enough. The pricing policy has to answer a harder question: what is the best accessible source, and how do we justify it when it is not the most liquid one? Possible answers include a less active exchange with adjustments, on-chain liquidity, or the 1:1 conversion into USDS. Each one needs to be documented and defended.
Active market or not?
The second question is whether the chosen market is active at the measurement date, meaning transactions take place with enough frequency and volume to provide pricing information on an ongoing basis.
This drives the fair value hierarchy:
A quoted price in an active market for the same asset is a Level 1 input, used without adjustment.
A price from a market that is not active, or that needs adjusting, falls to Level 2 or below, and must be justified.
On 30 September at 14:00, Kraken's price was three minutes old and backed by continuous trading. Gemini's price was six hours old and backed by $1.90. Treating both as equally valid "closing prices" would put a stale, near-meaningless print on the same footing as a live market.
In practice, this is the difference between a price an auditor accepts without discussion and one that triggers questions about the source, the methodology and the controls in place.
Choosing the right market is necessary, not sufficient
Kraken was the right market. It was not a perfectly calm one.
At 13:54:59, five minutes before the cut-off, a sell order swept Kraken's book: 13 trades, about 12,400 DAI, down to $0.98019. Six seconds later, the price was back at $0.99982. A similar sweep had already pushed DAI to $0.9884 at 11:56.
On the same market, at the same cut-off, the "price" depends on the rule applied:
Last trade before 14:00: $0.99994
Last trade before 13:55:01: $0.98019
Volume-weighted average over the last hour: about $0.9871, pulled down by the 13:55 sweep
Gemini's last trade, for comparison: $0.99771, or $0.98 without the $1.90 trade
For a stablecoin, that is a range of about 2%. The market choice removes the worst sources of noise. The rules applied on that market (staleness, minimum size, treatment of sweeps) handle the rest. Both need to be written down in the pricing policy, before the cut-off, not argued about after.
Why this matters for a fund
Stablecoins often sit in the "cash" bucket of a crypto fund. They rarely get the scrutiny applied to the rest of the book, precisely because they are supposed to be worth $1.
A 2% difference on a stablecoin is not small in NAV terms. In Luxembourg, CSSF Circular 24/856 sets the NAV error tolerance at 1% for retail funds invested mainly in shares or other assets. Professional funds such as SIFs set their own threshold. A fund holding half of its NAV in DAI would see a 1% NAV difference from the choice of market alone.
To be clear: choosing a price source in line with a documented pricing policy is not a NAV error. The point is the order of magnitude. A decision that is often made implicitly, by default or by habit, can move the NAV by as much as the threshold regulators use to define a material error.
Six questions for your pricing policy
How is the principal market selected? On what volume, over what period, and is the selection reviewed regularly?
Does the fund have access to that market? The principal market is assessed from the fund's point of view: accounts, custody, counterparties.
How old can a trade be and still count as a price? Is there a maximum age, and what happens when it is exceeded?
What is the minimum size of a representative trade? Would a $1.90 trade be accepted as a closing price?
How are sweeps and outliers treated? Is a trade that is reversed within seconds a price, or noise?
What happens when an asset is being wound down? Delistings, migrations (DAI to USDS) and regulatory restrictions (MiCA) can remove the principal market overnight. Who monitors this, and what is the fallback?
If the answers are not written down, they are being made implicitly at every cut-off, usually by whatever the data provider returns by default.
How we approach it at Koinju
We select the principal market for each asset on its traded volume over the previous 30 days, then build the price from that market's trades at the measurement time. Every price comes with its full explanation: the candidate markets, their volumes, the market retained, and the trades used. Anyone can trace the number back to its source and reproduce it.
On 30 September, that process retained Kraken for DAI/USD, at $0.99994. For a fund without access to Kraken's DAI market, the analysis has to be redone on the markets it can actually reach.
The principal market selection is a starting point. Each fund still needs to confirm that it has access to that market and that the result fits its own pricing policy.
Conclusion
The September close on DAI shows that the hardest part of crypto pricing is not the formula. It is the choice of where the price comes from, and the rules that turn trades into a price.
On a stablecoin, on two regulated exchanges, the difference between a live market and a stale one was a $1.90 trade from the morning. And for a European fund, the live market may not even be accessible. On a less liquid asset, the gap would be much wider.
Call to action: try the Koinju API for free here
For information only. This article does not constitute a Fair Value Assessment or investment advice. Data: Koinju trade-level data, DAI/USD on Kraken and Gemini, 31 August to 30 September 2026.
