How to Become a Crypto Market Maker: Capital, Infrastructure, Venue Access and Licensing

A market making desk sells immediacy. It posts a bid and an ask, earns the spread and any maker rebate when both sides trade, and loses money when the market moves against the inventory it has just accumulated. Everything else — the technology, the risk limits, the venue relationships — exists to make that basic trade profitable more often than not. Profitability is a function of spread captured, fill rate, rebate tier, and how efficiently the desk hedges the inventory it did not want.

Crypto differs from equities in four ways that dominate the build. Markets never close, so you need genuine 24/7 coverage and automated failover rather than a trading floor. Venues are fragmented and unequal, so reference pricing and cross-venue hedging matter more than raw latency in most pairs. Settlement and custody are your problem, not a clearing house's. And counterparty risk is real: capital parked on an exchange is exposed to that exchange.

The floor depends on scope. Quoting a single mid-cap pair on one venue with modest size can be attempted with low six figures of working capital, but the economics are marginal after fees and the desk has no cushion for an adverse inventory event. Running a handful of pairs across three or four venues with credible depth realistically starts in the low seven figures, because capital must be pre-positioned on each venue simultaneously — you cannot net across exchanges in real time, so idle inventory is a structural cost, not an inefficiency.

Add a reserve. The failure mode that kills new desks is not a bad strategy but an inventory shock during a volatility event, where the hedge slips, the book is one-sided, and there is no capital left to keep quoting. Desks that survive size their per-pair inventory limits so that a two-standard-deviation move consumes a defined and survivable fraction of equity, and they enforce those limits automatically rather than by judgement at three in the morning.

The core stack is a market data layer normalising each venue's feed, a pricing engine producing a reference mid, a quoting engine posting layered orders with dynamic spread and size, an order management layer handling rate limits and rejects, a hedging module, and a risk service with hard kill switches. Above that sits monitoring and reconciliation, because in crypto your position of record is whatever the exchange says it is, and drift between your internal ledger and venue balances is a daily reality.

Latency matters less than founders expect outside the top pairs. In BTC and ETH perpetuals on major venues you are competing with co-located firms and the arms race is real. In a mid-cap altcoin book, the binding constraints are inventory management, hedge availability and rate limits, not microseconds. Colocation in the venue's cloud region — AWS Tokyo for several major exchanges — is usually sufficient. Build for reliability and observability before speed: an engine that quotes correctly 99.9% of the time beats one that quotes fast and wrongly.

Every major exchange runs a market maker programme with fee tiers that turn negative — you are paid a rebate to post maker orders — once you meet volume, spread and uptime obligations. Access is granted per account, usually via an application that asks for firm details, AUM or trading capital, trading history, technical setup, and the pairs you intend to quote. Institutional accounts require full corporate KYC: certificate of incorporation, ownership chart to ultimate beneficial owners, directors' identification, source of funds, and an AML policy for the entity.

Programme obligations are measured in the same units you will later be measured on by issuers: time-weighted spread within a target, minimum depth at defined bands, and two-sided uptime across a monthly window. Fall below and you drop a tier, which can flip the desk's economics from positive to negative overnight. Xavion's institutional access work sits here: getting an entity through corporate onboarding, into the right fee tier, and connected to the right desk contacts is a relationship exercise as much as a paperwork one.

Proprietary market making — trading your own capital, holding no client assets — is treated very differently from managing outside money or providing a regulated service. The moment external capital enters, you are likely in fund or asset-management territory and the jurisdiction and licensing analysis becomes the first question, not the last. The same applies if you intend to quote to clients bilaterally rather than only on exchange order books.

Practically, a desk needs an operating entity in a jurisdiction that exchanges will onboard, banking that tolerates high-frequency fiat movement and does not treat turnover as an AML anomaly, and a documented compliance framework covering AML, market conduct, wash trading prohibition, and conflicts where the desk also holds issuer mandates. That last conflict is where reputational failures cluster, and where written policy separating principal trading from designated market making earns its keep. We advise on the structure, the banking rails and the venue onboarding as one connected problem, because solving them separately is how desks end up with an entity no bank will serve.

Almost never at first. A treasury quoting its own token faces a governance and market-conduct problem before it faces a technical one: the issuer controls both supply information and the book, which is exactly the conflict regulators and exchanges scrutinise. Even where it is permissible, the cost of building a reliable 24/7 desk exceeds several years of a retainer with an external firm, and the reputational downside of getting it wrong is asymmetric.

There is a narrower version that does make sense: an issuer-run treasury operations function that measures the external market maker rather than replacing it. Pull the venue data yourself, compute time-weighted spread, depth at each band and uptime independently, and hold the provider to the annex. That is a modest engineering effort with a direct financial return, and it is what serious foundations do.