Last year, more than 11.5 million tokens stopped trading. The CoinGecko chart runs 2021 through 2025 and the line bends one direction only. Most of the corpses were memecoins. Pump.fun lets anyone mint one in minutes for the price of gas and a click. Scarcity was the first thing this industry killed. Nobody held a funeral.
I pulled that chart on a Tuesday, cross-checked it against my own wallet history, and found the same pattern. About 90% of the tokens I held in 2021 no longer have a bid. Not a crash. A disappearance from the order book entirely.
That matters, because the wallet business is built on a lie that used to be true: that listing an asset is a competitive act. In 2018 I chose my first self-custody setup because it supported a token the others didn't. That edge is dead. Any wallet can integrate any asset in an afternoon. The listing list is a commodity now.
The industry has already written the sequel. The pitch goes like this: users don't want storage anymore, they want to swap, bridge, stake, earn on stablecoins and spend from a card, all inside one self-custody app. Building that stack yourself is too expensive, so wallets should plug into third-party exchange APIs, card processors and bridge aggregators. The result is what the marketing calls the "action wallet" — and the people selling that vision insist the winner will be measured by usage frequency, not token count.
The direction is right. The evidence backing it is weaker than the pitch, and I want to show you exactly where it breaks.
The structure of the claim
The logic chain runs in seven steps: token issuance is infinite, so tokens are no longer scarce; any wallet can list the same asset after one integration, so listing is commoditized; users want one-click buy, bridge, stake, stablecoin yield and card spending under self-custody; building swaps, compliance, cards and bridging is hard and needs licenses and years; therefore wallets should buy those components from third parties; therefore the wallet evolves from a storage app into an action app, earning on swap spreads, stablecoin flows, staking and card volume; the next winner is defined by usage frequency.
Steps one through three hold up. Token long-tail commoditization and demand for an action entry point are consensus by now.
Step four is where the argument starts to lean. "Hard" is not the same as "must be outsourced." Large wallets build in-house. Mid-size wallets aggregate multiple liquidity sources. Small wallets white-label. These are three different strategies, and the pitch collapses them into one.
Steps five and six contain the jump. The piece treats third-party API integration as the realistic path. It is a path. It is not the only one, and it happens to be the one the author's sponsor sells.
Step seven is directionally correct but sloppy. Frequency is not a moat. Retention, trust, distribution, compliance and capital efficiency are the moats. Frequency is the outcome when those work.
Here is the tension the pitch never resolves. If listing commoditized wallets, then swap, bridge, staking and card features will commoditize them too. The wallet stops being an asset-list front-end and becomes an API front-end. Value capture migrates upstream — to liquidity aggregators, compliance rails, card networks, intent solvers. The pitch says the wallet becomes an action wallet. It never answers what makes one action wallet different from another once everyone plugs into the same pipes. Its answer is "move early, lock good partnerships." That is supplier relationship advice, not a durable moat.
What an action wallet actually is
Strip the marketing and five capability layers remain. Asset layer: multi-chain, multi-asset, self-custody keys. Transaction layer: built-in swaps, cross-chain routing, aggregated liquidity. Yield layer: staking, stablecoin yield. Payment layer: on and off-ramps, card spending, day-to-day stablecoin payments. Experience layer: account abstraction, gas sponsorship, batching, permissions, hiding the chain entirely.
The framework is clean. The error is conflating "has every feature" with "a user will come back tomorrow." The variables that actually decide wallet winners are less glamorous:
- Whether fiat on-ramps clear without a support ticket.
- Swap slippage, failure rate, and price against a reference DEX.
- Security incidents and whether the brand survived them.
- Localization and distribution in the markets that matter.
- Whether the stablecoin card is accepted at a real merchant.
- Whether compliance and risk teams can sustain the setup under pressure.
- Whether the user returns without a notification.
So the precise definition should be: a self-custody wallet that completes high-frequency money actions with low friction. Not the wallet with the most features. Feature count is a spec sheet. Friction is the product.
The case data problem
Now the part that should make any analyst uncomfortable.
The flagship examples are D'CENT and Cake Wallet. D'CENT, a Korean hardware wallet, integrated the ChangeNOW exchange API, added in-app swap, and the piece reports one million registrations, forty thousand daily active users and hardware sales nearly tripling. Cake Wallet went from a single-asset app to multi-asset, integrated the same exchange API, and the piece reports growth from fifteen thousand downloads to 1.75 million users with three hundred thousand monthly actives.
Both use the same vendor. Both are cited in a piece that embeds that vendor's official posts and quotes its CBDO. That does not make the numbers false. It makes the sample selected, and it makes the causal claim unverified.
Start with units. Fifteen thousand downloads versus 1.75 million users is not a 117x comparison. It is two different metrics wearing the same jacket. A download is an install event. A user is a person attached to a key. If 1.75 million is cumulative, then 300,000 monthly actives puts the ratio near 17%, which is respectable for a consumer wallet and tells you nothing about whether the API integration caused it. Cake's reputation was built on privacy and community distribution. Those are the variables I would test first, not the swap widget.
On D'CENT: forty thousand DAU against one million registrations is a 4% daily engagement rate. Whether that is healthy depends entirely on whether the million is cumulative lifetime or annual additions, and the piece does not say. Hardware sales nearly tripling could be explained by a bull tape, Korean retail demand, or category-wide hardware recovery. There is no stated control window, no mention of simultaneous marketing, no pre/post comparison that isolates the integration.
There is also a definitional problem in the headline token stat. "More than 11.5 million tokens died" — died how? Zeroed out, delisted, or simply stopped trading? The meme-coin share is asserted, not measured. And the stablecoin claim — trillions in settlement volume, approaching card networks, sub-second and sub-cent — carries no source at all. On-chain settlement volume includes internal transfers and bot traffic. Comparing it directly to card network throughput is apples to a spreadsheet.
The one genuinely verifiable technical detail is MetaMask's scoped permission model, something like a hard cap of 10 USDC per day inside a thirty-day policy. That is real progress on permissions and worth watching. It also has nothing to do with API integration, which tells you how thin the connective tissue is.
Where the pitch is structurally wrong
Here is the angle the vendor content will not give you. The "API makes you an action wallet" thesis assumes the API is neutral plumbing. It is not. Every third-party integration introduces counterparty risk the wallet's users never consented to and the wallet's brand absorbs entirely.
Four risks never appear in the pitch:
Vendor lock-in. Once your swap routing, fee logic and UI flows are wired to one provider, migrating is a quarterly engineering project, not a config change. Fee changes and service degradations become non-negotiable.
Privacy leakage. A non-custodial wallet that routes every swap through a hosted API is advertising every trade to a third party in real time. For a wallet whose brand is privacy, that is a category error.
Liquidity opacity. "Best route" is a marketing string until you measure realized execution against a reference venue. My own logs from DeFi Summer taught me the same lesson the hard way: the alpha was in the code, not the community hype. Route quality is measurable. Most of these integrations don't publish it.
Conflict of interest in revenue. Swap rebates, card cashback and staking referral fees all incentivize frequency. A wallet paid per swap has a structural incentive to route through the venue that pays it, not the venue that executes best. The user pays the difference in slippage.
And the deeper strategic problem: if every mid-size wallet plugs into the same exchanges and card rails, the differentiated surface shrinks to brand, distribution and fiat connectivity. The feature layer commoditizes exactly the way the listing layer did. The pitch's own logic eats its conclusion. The chart does not lie, only the ego does.
What I am actually watching
I do not trade wallets. I trade flows, and wallets are where retail flow first becomes visible. So I watch the metrics that precede price, not the ones that decorate a pitch deck.
Four things, in order of signal quality. First, on-ramp conversion — how many users who start a fiat deposit finish it. That number is the real product. Second, swap execution quality — realized slippage against a reference DEX for a standardized size, published by the wallet, not the vendor. Third, stablecoin balance retention over 30 days — not volume, balance. Volume can be bots. Balance is intent. Fourth, dependency concentration — how many wallets route through the same single API. When that concentration crosses a threshold, you are not looking at a competitive market. You are looking at a single point of failure wearing twelve logos.
For the teams: stop reporting supported token counts as a KPI. Report 30-day return rate, swap success rate, ramp completion, and revenue per retained user. Run multi-provider routing so no single API owns your execution. And treat compliance as a first-class product — non-custodial does not mean unregulated once you touch cards and fiat. The value is in trust, distribution, localization and capital efficiency. Yields are signals; liquidity is the only truth.
The genuine shift is real: wallets are moving from asset coverage to action frequency, and stablecoins plus account abstraction make daily use possible. The weak link is the causality — a vendor-shaped case study is a lead, not a proof. Plugging into an API accelerates a small wallet's roadmap. It does not make it a moat, and it may quietly turn the wallet into someone else's front-end.
So the question worth asking is not which wallet lists the most tokens. It is which wallet you would still open on a day when the API is down, the card is declined, and the chart is red — and how fast the others route you out.