The error is mechanical and anyone can check it inside five minutes. That is what makes it embarrassing rather than interesting. Bullish and Equiniti, four point two billion, published as forty two. Mastercard and BVNK, up to one point eight billion, published as eighteen. Figure and Kiavi, seven hundred and seventeen million, published as seven point one seven billion. IREN and Mirantis, six hundred and twenty five million, published as six point two five billion. Payward, the parent of Kraken, and Reap, up to six hundred million, published as six billion. I went through the list line by line. Every entry had moved one decimal place to the right, so the total moved with them.

I want to be precise about what went wrong, because nobody invented these transactions. They all took place as described. The targets are real, the acquirers are real, the announcement dates are real. What moved was the unit. Once a unit slips inside a table it travels into the headline, and a wrong headline outruns its correction every time. I have now seen the ninety three point seven figure quoted in two decks and one newsletter. Not one of the three cited a primary register.

CryptoRank Research tracked 87 acquisition announcements in the crypto sector between January and June 2026. It recorded $9.66 billion in disclosed value, and that is the record everyone is reaching for. I think it is more interesting than the fake one. The figure beats the second half of 2025 by 223% on value, while the deal count fell 25%, from 116 announcements down to 87. Fewer transactions carrying much more money is the actual shape of the half.

The quarterly split inside that total is the part I would hold onto. Q1 2026 carried 46 announcements and Q2 carried 41. The peak on count sits back in Q4 2025 at 61, so the count has fallen for three consecutive quarters while value climbed. Anyone reading the inflated total would conclude the market is broadening out. The real numbers say the opposite.

One transaction supplied 43% of the entire half. Bullish and Equiniti at $4.2 billion sits against a total of $9.66 billion. The top four deals together supplied 76% of the disclosed value. Add Mastercard and BVNK at up to $1.8 billion, Figure and Kiavi at $717 million, and IREN and Mirantis at $625 million. Four line items and I have accounted for three quarters of a record half.

When I strip those four out, the remaining 83 announcements share about $2.3 billion between them. That is a normal half by any reading, and arguably a quiet one. The record is one deal and three follow-ups. I would not describe that as a market-wide repricing in front of anyone who divides.

This is the number I would put in front of any owner about to price an asset off a headline. The median disclosed crypto transaction in H1 2026 stayed flat at $100 million. That is exactly where it sat in the previous half year, while the aggregate more than tripled.

A median that holds while the total triples tells me the top of the market moved and the body of it did not. If you are selling an EMI, a small exchange, a payments licence or a custody stack, I would anchor on the median. I have watched three owners this year anchor on an aggregate instead. Each of them spent about four months discovering that buyers do not price off aggregates.

Only 24% of the 87 announcements disclosed a value. Sixty six deals closed with no public number attached to them. Every aggregate anyone quotes, including the correct $9.66 billion, therefore describes roughly a quarter of the activity. It says nothing whatever about the rest.

I think that gap is the biggest single reason licensed-asset pricing stays opaque. A buyer walks in with four comparable transactions, three of which were never priced publicly. The seller has no way to test the fourth. The N5Deal Asset ID Card exists for that gap. It presents the licence, the permissions, the jurisdiction and the operating history in a structured form, so the conversation starts from the asset's documented facts. It does not manufacture comparables out of nothing, and I would not claim that it does. What it stops is a negotiation running on a number neither party can source.

The sector split of the 87 deals runs as follows. Infrastructure took 19 deals and 22% of the half. DeFi took 9 at 10.3%, exchanges took 7, analytics platforms took 7, and payments and social took 5 each. Infrastructure therefore took more deals than DeFi and exchanges combined.

I read that as the real signal of the half, and it is the one buried under the fake total. Money is moving into settlement, custody, registers and rails, not into consumer-facing protocol brands. Bullish did not buy a trading app with its $4.2 billion. It bought a transfer agent, the unglamorous regulated entity that maintains the record of who owns what. It bought that in order to become the register for tokenised securities.

KPMG published its Pulse of Fintech for H1 2026 on 26 August 2026, with a 30 June cutoff. Total global fintech investment came in at $103.1 billion across 2,100 deals. The comparison is $72.2 billion across 2,501 deals in H2 2025. I notice the same pattern as in crypto: value up 42.8%, count down by 401 deals.

Inside that report KPMG puts digital assets and crypto at $11.1 billion across 467 deals. CryptoRank puts crypto M&A at $9.66 billion across 87. Both are right and I would not place them side by side. KPMG counts venture, private equity and M&A together across a wider sector definition. CryptoRank counts acquisition announcements only, and only the quarter of those that published a price. The denominator is chosen rather than measured. Anyone quoting one figure against the other is answering two different questions at once.

Payments took $44.2 billion of the KPMG total across 168 deals, more than four times the entire digital assets line. AI-focused fintechs took another $21.4 billion. The two largest transactions of the half were Global Payments buying Worldpay at $24.3 billion and the TSYS divestiture to FIS at $13.5 billion, both announced in January 2026. Either one alone is larger than every crypto acquisition of the half combined.

When I look at the four deals that made the crypto record, the object of the purchase is not technology. Equiniti is a transfer agent with a regulated register. BVNK is a stablecoin payments business with the licensing surface that implies. Reap is a payment gateway and Kiavi is a lender.

Each target carried permissions, a supervised entity, or a book of clients the buyer could not have built inside a reasonable calendar. That is what I believe the money bought. The code in most of these businesses could be rewritten by a competent team inside a year. The authorisation behind them could not be obtained inside a year in any of the jurisdictions involved. The premium sits on the regulatory side of the balance sheet. I find the H1 distribution says so more clearly than the commentary written around it.

I would take three things from it and none of them are comfortable. Your asset is priced against the $100 million median, not against the $9.66 billion aggregate. The exception is an asset carrying a register, a licence set or a client book a strategic buyer treats as unbuildable. The number of buyers is falling, since deal count has dropped for three straight quarters. I would treat a slower process as the base case rather than the bad case. And the 24% disclosure rate means your comparables are mostly invisible, so the side with better documented facts controls the pricing conversation.

That last point is the one I keep returning to. The buyer in a licensed-asset transaction has usually seen twenty processes and the seller has seen one. I would use the N5Deal KYF Intelligence layer and the real-time multiples drawn from platform deal flow to narrow that asymmetry. It matters far more in a market where three out of four transactions never publish a price.

I could not establish the undisclosed value of the 66 crypto deals that closed without a price. No tracker I read publishes an estimate on a basis I would defend. If the median of those deals matched the disclosed median, the half would total near sixteen billion dollars. I want to be clear that this is arithmetic rather than evidence. Undisclosed deals skew small, everyone in the market believes it, and I could not find the belief demonstrated on any published dataset.

I could not establish who first published the ten times figure. I found it on a compliance-focused site citing CryptoRank and BitcoinWorld for a June 2026 aggregate. The error is not present in CryptoRank's own report, so the multiplication happened somewhere downstream. I have not traced the first instance. I would rather say that than name a publisher I cannot prove is the origin.

I also could not reconcile the KPMG digital assets figure with the CryptoRank figure on any line-by-line basis. Neither party publishes the underlying deal list in full. The two sit close on value and four hundred deals apart on count. That tells me the definitions of a crypto transaction differ substantially between them, and I cannot say by how much.

One detail still bothers me. The largest crypto acquisition of the half bought a transfer agent, an institution most people in this sector could not have defined a year ago. I have not read a single piece of commentary that treats that as the story.

This article is market commentary on fintech and crypto M&A activity in the first half of 2026. It is not legal, financial, tax or investment advice on a specific transaction, asset or licence. Deal values, counts and sector classifications differ between trackers according to what each one counts and when. Disclosed values cover only a minority of transactions. Verify every figure quoted here against the current release of the underlying source before using it in a valuation, a board paper or a marketing claim. Before pricing, marketing or acquiring a regulated financial asset, take the specific case to a transactions lawyer in the relevant jurisdiction, to an M&A adviser with experience in that licence category, and to a regulatory consultant who has taken a change of control application through the supervisor in question