Crypto OTC trading: how block desks actually work

What an OTC desk actually is

An exchange works by matching your order against a public book of resting orders. That is efficient for ordinary size and terrible for large size, because a big order eats through the book, each successive fill worse than the last, and everyone watching can see it happening. By the time you are done you have paid an average price meaningfully worse than the one you saw when you started, and you have told the whole market what you were doing.

An over-the-counter desk solves that by taking the trade off the book entirely. You ask for a price on a specific size, the desk gives you one number, and you either take it or you do not. If you take it, the whole trade happens at that single price. Nothing appears on any order book, there is no partial fill, and there is no slippage, because the price was agreed before anything moved.

The other half of the definition is who these desks will deal with. They are institutional businesses and they say so explicitly — B2C2 publishes that it "does not transact with or provide any service to any retail investor or consumer", and Crypto.com describes its desk as being for selected institutional and VIP clients. Onboarding is corporate know-your-customer: entity documents, beneficial ownership, source of funds, sanctions screening and wallet screening. It takes weeks, not the minutes an exchange signup takes.

Once you are onboarded, you either prefund — sending cash or coin to the desk before you can trade — or you negotiate credit. Prefunding is the default everywhere. Credit is a relationship decision, and the desks that publish it are specific: Kraken Prime advertises T+1 credit facilities and asset-backed lending, and CEX.IO's institutional API exposes credit line and overdraft fields.

The RFQ, second by second

The request-for-quote flow is the heart of it, and the timings are tighter than most people expect. You send the side, the pair and the size. The desk returns a firm price. You have a few seconds to accept it.

Coinbase Prime is the most specific about this of any desk in the market, and its published numbers are worth quoting exactly: the price hold is three seconds on the interface, refreshable up to twenty times, and one second through the API. The order is fill-or-kill, it never touches the central order book, and the quote is returned inclusive of all fees and commissions. Crypto.com publishes a minimum ten-second window on its client portal, with longer windows available by voice or API. Kraken says only that you receive executable quotes "in seconds", with no obligation to trade, and puts no number on it.

Those windows exist because the desk is quoting a firm price on an asset that moves. Every second it leaves that quote live, it is short an option to you. Three seconds is not arbitrary — it is roughly what a market maker can afford to give away on a volatile asset before the optionality costs more than the trade earns.

Principal or agency — and it is not uniform

This is the most consequential thing on the page and the point most consumer explanations get wrong. They tend to state flatly that an OTC desk trades as principal. Some do. The largest one does not.

Coinbase Prime publishes agency-only execution — its own wording is that "agency-only execution and transparent post-trade reporting ensure client interests come first". Its RFQ is not Coinbase quoting you a price at all; it is an auction among third-party liquidity providers, from which the winning quote is passed through. Coinbase is not the risk-taker.

Galaxy publishes the opposite, and treats it as the selling point. When it launched institutional over-the-counter prediction-markets trading in June 2026, its global co-head of trading described the offering as giving clients "a principal counterparty that can warehouse risk". Market makers such as B2C2, Wintermute, Cumberland, GSR, Keyrock and Flowdesk are principal by construction — quoting their own book is what they do. Gemini runs both models side by side: a voice desk plus an electronic service priced from external liquidity providers.

Why it matters to you: with a principal desk, your cost is the spread and your counterparty is the desk itself, which has taken the other side and now has an interest in that position. With an agency desk, your cost is a negotiated commission and the desk is routing rather than risking. Neither is better in the abstract. But you should know which one you are dealing with, and a surprising number of clients do not.

Block price versus worked order

There are two ways to move size, and choosing between them is the main judgement a desk client makes.

A block is the firm risk price described above. You get certainty immediately: one price, one fill, done. You pay for that certainty in the spread, because the desk is taking on a position it now has to manage in a market that may move against it.

A worked order is the alternative. You ask the desk to execute gradually across venues over hours, tracking a benchmark — time-weighted average price, volume-weighted average price, or implementation shortfall. You pay a commission in basis points rather than a spread, and you accept execution risk in exchange for a tighter expected average. Coinbase Prime publishes unusually specific parameters for this: its TWAP and VWAP algorithms break an order into buckets with a $100 minimum notional per bucket, its market orders are submitted as immediate-or-cancel limits with a 5% slippage collar capped at five seconds, and its stop-limit orders require the limit price to sit within 100 basis points of the stop.

The clean way to think about it: the block price is an insurance premium against the market moving while you execute. Sometimes that insurance is cheap and worth buying. In a quiet market, working the order will usually beat it.

Execution

Minimum ticket sizes — the received wisdom is out of date

You will read almost everywhere that crypto OTC starts somewhere between $50,000 and $250,000. That was true. It is now only half true, because electronic request-for-quote has pushed published minimums dramatically lower while the six-figure numbers survive on voice-desk marketing pages.

The reason for the shift is competitive rather than charitable. Electronic RFQ costs a desk almost nothing per trade, so Coinbase, Gemini and Crypto.com have all pushed their thresholds down to capture flow that would otherwise go to their own order books at a lower margin. The practical consequence for you is that "OTC" no longer describes a size band. It describes an execution method.

That said, the economics still bite at the bottom. Below roughly $50,000 a retail buyer is almost always better served by a limit order on a deep exchange book, because at that size the desk's embedded spread will exceed the exchange's taker fee plus whatever slippage you would actually have suffered. Our guide to order types covers how to place that limit properly.

What a desk charges, and why you cannot look it up

Here is the honest position, stated plainly because the alternative is to make something up: no crypto OTC desk publishes its spreads, and there is no credible public dataset measuring realised OTC spreads. Any article giving you a confident basis-point figure for crypto OTC is extrapolating at best.

What can be verified is the pricing model, and it is consistent across the industry. Desks quote an all-in price with their economics inside it. Kraken's wording is the clearest: "we charge no fees for our service — the quoted price you see is the whole price". Crypto.com states there are no trading fees on its OTC desk. Gemini says there are no additional fees to trade or settle on its electronic service. All three statements are true, and all three describe the invoice rather than the cost. The desk is paid in the price.

Coinbase Prime goes furthest of anyone, publishing the structure while withholding the number. It documents that trading fees are applied "using a negotiated basis points rate", and that clients can be on either all-in pricing, where one consolidated commission appears, or cost-plus pricing, where the commission and the exchange fee are shown separately. It then states outright that the fee amounts in its own examples are illustrative and do not reflect the rates a client would actually receive. That is unusually candid, and it tells you the thing worth knowing: this is a negotiated market and your rate depends on who you are.

Directionally, one dataset is worth citing because a desk publishes its own book. Wintermute reports that institutions accounted for 72% of its spot OTC volume in the first half of 2026, up from 59% a year earlier, and that bitcoin and ether fell from 54% of notional volume in 2023 to 49% in 2025. Both point the same way: pricing in the majors is getting more competitive as more institutions compete for it, while the long tail is getting wider and more dispersed. That is a direction of travel, not a number you can put in a spreadsheet.

Settlement, and the change that mattered most

Agreeing a price is the easy half. Actually exchanging the assets is where the risk lives, and it is where this market changed most after 2022.

The tri-party model in that last row is the important one, and it deserves explaining properly because it is the single most significant structural change in institutional crypto since the exchange failures of 2022.

Zodia Custody describes the mechanism as clearly as anyone: assets sit "in a fully segregated institution-grade trading wallet, so they remain your property, completely isolated and protected until trade settlement", and the network "enables participating venues to establish limits based on your assets mirrored on a wallet held by Zodia Custody, thereby eliminating the need to pre-fund".

Three consequences follow. The exchange never takes title to your assets — it extends you a trading limit against a mirror of a balance held at a regulated custodian. Pre-funding disappears, which collapses your worst-case exposure from "everything I have on the venue" to "whatever is unsettled since the last cycle". And the same collateral can support positions across several venues at once instead of being fragmented and stranded, which is the commercial reason the model spread so quickly.

What it does not do is eliminate counterparty risk, and anyone telling you otherwise is selling something. Settlement is periodic and net, so the delta between cycles is real exposure. Tri-party bounds the risk; it does not remove it. The main networks are Copper ClearLoop, Fireblocks Off-Exchange, BitGo's Go Network and Zodia Interchange. BitGo and Copper connected their networks in February 2025 in what both described as a first of its kind, letting clients hold assets with BitGo Trust and trade on partner venues without prefunding — notable mainly because until then each network had been its own silo.

Who runs the desks

The list below is not exhaustive and deliberately excludes anyone whose current status we could not verify. Two entries need care because they are widely misdescribed.

Genesis should be spoken about in the past tense. Genesis Global Capital filed for Chapter 11 in January 2023 after the Three Arrows Capital and FTX contagion, and Genesis Global Trading — the OTC desk itself — ceased operations in September 2023, surrendering its New York BitLicense and paying an $8m penalty in a settlement with the state regulator. What remains is litigation: the estate's litigation oversight committee filed suits in May 2025 against Digital Currency Group and its insiders, seeking at least $2.2bn in Delaware and more than $1bn in a separate clawback action. It is not a desk you can trade with.

Talos is not a desk at all. It is an execution and order management system that routes to desks and venues. It is frequently listed alongside liquidity providers, and it is a different kind of business entirely.

Others operating credibly in this space include GSR, Flowdesk, Keyrock, Amber Premium — now listed on Nasdaq — and Zerocap in Australia. Consolidation has been the theme of the last two years: Coinbase bought Deribit, Kraken bought NinjaTrader, Ripple bought Hidden Road, FalconX bought Arbelos Markets and Robinhood bought Bitstamp. The independent middle of this market is thinner than it was.

Do you actually need one?

For the overwhelming majority of readers, no. The honest test is whether your order would visibly move the book you are trading on. If you are buying $5,000 of bitcoin on a major exchange, it will not, and a desk relationship would cost you weeks of onboarding to obtain a worse price. Our page on where to buy crypto covers the retail venues properly, including the point that on nearly every major exchange the professional order book is one tab away from the beginner interface and costs a fraction as much.

Where a desk earns its keep is at the point your order stops being a price-taker and starts being a price-mover. If you are deploying several hundred thousand dollars or more, the market impact you avoid will comfortably exceed the spread you pay, and the certainty of a single fill at a known price has real value in an asset that can move several per cent in an hour. High-net-worth individuals, family offices, treasury operations at companies holding bitcoin on the balance sheet, and funds rebalancing are the genuine use cases.

There is also a third option people forget. If what you want is bitcoin exposure rather than bitcoin coins — inside a portfolio, in a retirement account, alongside equities — then none of this applies. A spot Bitcoin ETF gives you that through a broker you probably already have, with no onboarding, no custody decision and no counterparty relationship to negotiate. The trade-off is explained in full on ETF versus owning bitcoin. Large ETF orders have their own version of this problem, incidentally, which is why authorised participants exist at all — we cover that machinery in how Bitcoin ETFs work and name the firms involved on the issuers and custodians page.