Arcus Opens the Self-Custody Gate, but the Tolls Flow Elsewhere

There is a new place to trade Apple, Tesla, and Bitcoin around the clock, hold the assets yourself, and sign in with nothing more than an email address. It is called Arcus, it comes from the team behind the crypto derivatives platform dYdX, and it is built with Robinhood on Robinhood's own blockchain. The strategy hiding inside it is one sentence: the reason ordinary investors never adopted crypto self-custody is that the front door is awful.

That view belongs to Arcus's CEO, Eddie Zhang, who put it bluntly in July: the sign-up screen is the primary competitor. For most of crypto's history, buying in meant downloading a browser extension, saving a seed phrase, and trusting yourself not to lose it — a gauntlet that chased people back to the convenient brokerage. Arcus's answer was to buy the door down. It partnered with Privy, a wallet-infrastructure company, so that a self-custodial account appears silently behind an Apple- or email-login, with private keys generated in a secure enclave and recovery handled by simply signing in again. Users hold their own assets; it just no longer feels like work.

That is the gate in "self custody at the gate," and removing it is genuinely the hard part of the business — which is exactly why the economics of the deal deserve a second look. Because the money, it turns out, does not live at the gate at all.

Since launching in early July, Arcus has moved roughly $2 billion in cumulative trade volume, the bulk of it in perpetual futures. It kept almost none of it. Spot trading of its 95 tokenized stocks is fee-free; fees come only from perpetual futures, at a take rate of roughly 1.5 basis points. Over the trailing 30 days that produced about $247,000 in fees — annualized, near $4.5 million — all resting on about $19 million of deposits. A protocol can look busy and stay unprofitable, and Arcus is showing how: enormous flows, a thin toll, and deposits a fraction of the volume passing through.

Two additional facts matter for anyone in the United States. You cannot trade it — Arcus is not available in the U.S., Canada, or the U.K., and it has no brokerage or exchange registration to lean on. And there is no Arcus token yet, so there is nothing to buy but the expectation of one; the team has said any future token will reserve an allocation for the dYdX community. The closest public exposure has already moved: the existing DYDX token fell about 23% on the day the pivot was announced, as the market priced this as a separate company with its own token, its founder stepping onto the board.

The investment question is who actually captures value in this arrangement, and the Privy deal is the tell. Arcus correctly identified the gate as its constraint and then rented the solution rather than owning it. Privy is now part of Stripe, the default embedded-wallet layer for the industry, and Arcus pays a per-user infrastructure bill for the privilege of feeling frictionless. Distribution and the chain belong to Robinhood. What remains for Arcus is a thin application layer running 0% spot fees and a sub-two-basis-point derivatives toll, hoping a future token becomes the place where that flow finally lands.

None of this makes the wedge wrong. A fee-free, self-custodied market for tokenized equities and perps, with Robinhood's distribution and a frictionless login, is a promising attempt at a real gap: making crypto trading feel like a brokerage without surrendering the keys. But the tokenized stocks are contractual claims to economic exposure rather than ownership of the underlying shares, so the user also absorbs issuer and regulatory risk on top of the usual self-custody duty — the one-click login lowers the barrier but does not transfer responsibility.

What would prove this is a business rather than a promotion is measurable. Perpetual-fee revenue surviving the eventual airdrop, retail access to those perps, token terms that actually give the protocol a claim on the flow, and deposits growing toward the volume instead of floating far below it. On the current evidence — roughly $2 billion of volume against $19 million of deposits and a $4.5 million annualized fee run-rate — the gate has been removed, but almost all of the toll is still flowing to the user, to Stripe's wallet layer, and to Robinhood's rails. There is a coherent product here. What there is not yet is a clear answer to who, beyond the user, gets paid.