Crypto Exchange Development Cost: What 1,800 Hours Buys

The wallet plans that cost more than the plan below them

Three published pricing structures in this market invert the usual logic, and all three are computable from the vendors’ own rate cards. In each case, a plan marketed as the upgrade is the more expensive choice across a wide band of real usage.

Web3Auth: the free plan overtakes the $69 plan at 2,380 wallets

Base is free to 1,000 wallets, then $0.050 each. Growth is a flat $69 to 3,000 wallets.

Set them equal: 0.050 × (M − 1,000) = 69, so M − 1,000 = 1,380, and M = 2,380.

Check it: at 2,380 wallets, Base costs 0.05 × 1,380 = $69.00, and Growth costs $69.00. At 3,000 wallets, still inside Growth’s flat fee, Base costs $100.00 against Growth’s $69.00. The free plan is 45% more expensive than the paid one.

Web3Auth: the plan marked POPULAR is the wrong choice below 13,000 wallets

Growth is 69 + 0.045(M − 3,000). Scale is 399 + 0.040(M − 10,000). Setting them equal gives 0.045M − 66 = 0.040M − 1, so 0.005M = 65 and M = 13,000. Both cost $519.00 there.

The revealing point is 10,000 wallets, exactly where Scale’s allowance begins: Growth costs $384.00, and Scale costs $399.00. At the volume Scale is designed for, Scale is still the dearer option. It carries the POPULAR label anyway.

Magic: the free plan beats the “MOST POPULAR” plan to 8,800 wallets

Developer is 0.045 × (M − 1,000). Startup is 99 + 0.040(M − 2,500), which reduces to 0.040M − 1. Setting them equal gives 0.005M = 44, so M = 8,800, where both cost $351.00.

The plan Magic labels MOST POPULAR is worse than Magic’s own free plan for every customer below 8,800 monthly active wallets, which is 8.8 times the free tier’s allowance.

Why this keeps happening. In all three cases, the overage rate falls as the tier rises, from 5 cents to 4.5 to 4. Because the per-unit rate improves while the base fee jumps, the cheaper plan’s cost line stays below the dearer plan’s line well past the point where the dearer plan’s allowance begins. The structure is not a trick, but nothing on either pricing page tells you where the crossing is, and the crossing is the only number that matters.

What to do with this. Before signing, write your vendor’s plans out as base fee plus rate times volume, set two of them equal, and solve. It is one line of algebra, and it is the difference between choosing a plan and being assigned one.

What identity checks and money movement cost per transaction

The build is one-time, and the rails are permanent, so a quote covering only the build tells a founder little. Stripe publishes 2.9% plus 30 cents for domestic cards and 0.8% for ACH Direct Debit, capped at $5.00. Adyen publishes $0.13 plus interchange plus 0.60% on Visa and Mastercard. Neither appears on any ranking cost page.

At size, the caps are more important than the percentages. Stripe’s ACH cost is essentially stable over $625 per transaction, with a 0.8% cap at $5.00. GBP bank transfers are 0.5% capped at $5.00, and SEPA Direct Debit is 0.8% plus 30 cents capped at $6.00. Rail costs on large transfers will be significantly overstated by a model that applies the % without the cap, whereas small transfers will be understated by a model that ignores the percentage below the cap.

Adyen publishes its schedule with an explicit qualifier worth quoting before anyone builds a spreadsheet on it: The fees outlined above are indicative; please get in touch to discuss pricing options. Checkout.com publishes no rate at all, offering tailored pricing behind a contact form.

Exchange fee schedules, where they are published

For anyone modelling revenue rather than cost, Kraken and Binance publish complete spot fee schedules, and Coinbase does not.

Kraken’s schedule runs from 0.40% maker and 0.80% taker at the entry tier down to 0.00% maker and 0.10% taker above $10 million in 30-day volume, and it qualifies tiers on the better of volume or assets on platform. Binance runs from 0.100% maker and taker at the regular tier to 0.011% and 0.023% at VIP 9, with a 25% discount for paying fees in BNB.

Coinbase publishes no table. Its own help page states that to see the structure, a user must sign in to their Coinbase.com account and see the Coinbase Advanced fees page, and that Tiers update hourly based on trading volume. A competitor’s published schedule is a planning input. A gated one is not, and a business model built on assumed Coinbase parity is built on an assumption.

Identity verification is priced per check, and the definition of a check varies

Sumsub, Persona and Veriff publish per-verification pricing; Entrust, which now owns Onfido, routes to “Talk to an Identity Expert”. The number that moves a budget is not the headline per-check price but what counts as a check. A re-run after a failed capture, a document plus a liveness test, and an ongoing sanctions re-screen may each bill separately, and the per-check figure quoted in a sales conversation is usually the first of those.

Therefore, the practical challenge for a scoping document is not how much you charge for each verification, but rather what events result in a chargeable verification and how many are anticipated for each authorized user. Whatever the actual re-check rate is, a KYC line based on one check per user will be inaccurate.

White label or custom, and the break-even nobody computes

All nine pages ranking for white label crypto exchange vs custom are published by a party that sells one or both options. Zero are independent. Five take a side, and not one of the five discloses that it sells the side it recommends. Three disclose, and all three are vendors that sell both, which makes disclosure free.

The business model makes all of the recommendations. White-label is recommended for every page that sells exclusively white-label. Custom is recommended on any page that sells solely custom. Every page that offers both refuses to make a decision.

A disclosure before the method. Omega Solution sells custom development, so what follows is a way to compute the answer rather than the answer itself.

Five of the nine publish no dollar figures at all, on a decision that is purely economic. Zero of nine compute a break-even, which is the only question the query asks. Here is how to compute yours.

White-label costs a setup fee plus a monthly license, and often a share of revenue. Custom costs hours times rate, once, plus maintenance. The crossing point in months N is:

N = (custom build cost − white-label setup) ÷ (monthly license + monthly revenue share)

Worked with published figures rather than invented ones: B2Broker publishes B2TRADER starting from $2,500 / mo. CISIN, the one page on that SERP pricing both sides, publishes white-label setup at $30,000 to $80,000. Take Omega’s measured 1,800 hours at an Eastern European $90, which is $162,000, against a $30,000 setup and $2,500 a month:

N = (162,000 − 30,000) ÷ 2,500 = 52.8 months, a little over four years.

Change one input and the answer moves a long way. At South Asian rates, the same 1,800 hours is $36,000, and N = (36,000 − 30,000) ÷ 2,500 = 2.4 months. At B2Broker’s own higher published tier of $6,000 a month with an Eastern European build, N = 132,000 ÷ 6,000 = 22 months.

Which is why nobody publishes it. The answer is not a fact about the two options. It is a fact about your build rate and your license tier, and it swings from ten weeks to four years across published inputs. Any page that tells you white-label or custom wins without asking your rate is selling one of them.

One more term belongs in that division and is almost always left out: B2Broker states that A-book commissions offset all software subscriptions, meaning the license can fall to zero if you route order flow through the vendor. That is a real discount and also a lock-in, and it makes the monthly denominator a negotiation rather than a number.

What compliance costs, and when the license exceeds the build

Across the fifty-one pages Omega Solution audited, one prices regulatory licensing at all. Suffescom publishes US state money transmitter licenses at $50,000 to $200,000, against its own white-label MVP at $25,000 to $80,000. On its own figures, the license costs more than the software, and nine of the ten cost pages never mention licensing exists.

No central bank, statistics agency, or regulator discloses the cost of creating an exchange figure or the cost of obtaining a MiCA license. Each number that is in use for both originates from a company that offers the service. The clock, which is more helpful, is what regulators do publish.

The numbers that come from the regulation itself

MiCA is Regulation (EU) 2023/1114, and it has applied generally since 30 December 2024 under Article 149(2), with Titles III and IV running earlier from 30 June 2024 under Article 149(3). Article 59(1) is the operative sentence for anyone planning to serve EU users, and it has two limbs: A person shall not provide crypto-asset services, within the Union, unless that person is: (a) a legal person or other undertaking that has been authorised as a crypto-asset service provider in accordance with Article 63; or (b) a credit institution, central securities depository, investment firm, market operator, electronic money institution, UCITS management company, or an alternative investment fund manager that is allowed to provide crypto-asset services pursuant to Article 60.

Commercially speaking, limb (b) is often overlooked. A founder does not always require a new Article 63 permission if they already have an authorization for an investment firm or e-money institution under Article 60. Before scoping the compliance work, anyone purchasing or expanding on that premise should determine which limb applies.

Article 63 sets a statutory clock, not an estimate. The competent authority has 25 working days to assess completeness and 40 working days for the substantive assessment, with any suspension capped at 20 working days. A founder can plan against that. Consultancy estimates of six to twelve months cannot be checked against anything.

The transitional window is closed, and for many firms it closed earlier than the headline date. Article 143(3) is precise about this: providers operating lawfully before 30 December 2024 may continue to do so until 1 July 2026 or until they are granted or refused an authorisation pursuant to Article 63, whichever is sooner.

Read the last three words. A firm that applied and received a decision lost the transitional cover on the date of that decision, not on 1 July. Refusal ends it just as finally as approval. None of the pages this section corrects mentions that.

Member states could also shorten the window, and several did. On ESMA’s own list, Latvia, Hungary, the Netherlands, Poland, Slovenia and Finland ran six months and Sweden ran nine. ESMA attaches a caveat to that list worth carrying across: the periods were communicated by national authorities and may not all have been transposed into national law. Any page still advising founders that they can operate under a national regime into 2026 is describing a window that has shut.

As of 1 September 2026 the EBA’s own figures were 39 e-money tokens issued and zero asset-referenced tokens authorized. Content describing authorized ARTs under MiCA is describing something that has not happened.

The threshold that does not exist

This is the most common factual error in agency writing on the subject, and it is checkable in one sentence.

The EU travel rule for crypto is Regulation (EU) 2023/1113, on information accompanying transfers of funds and certain crypto-assets, OJ L 150/1 of 9 June 2023, in application since 30 December 2024. A great deal of published guidance states that it carries a €1,000 threshold for crypto. It does not. Recital 30 closes with this: In order to reflect those specific features, transfers of crypto-assets should be subject to the same requirements regardless of their amount and of whether they are domestic or cross-border transfers.

The €1,000 figure is real, and it lives elsewhere. It governs transfers of funds in Articles 5(2), 6(2) and 7(3), and it governs the ownership verification required on transfers to or from self-hosted addresses. Neither is a floor below which a crypto transfer escapes the traceability obligation.

Applying the funds threshold to crypto produces a compliance design that misses everything beneath it, which is the opposite of what the recital asks for. If a vendor’s scoping document contains a €1,000 crypto threshold, that vendor has taken the number from a summary rather than the regulation.