Crypto Exchange Development Cost: A Line-by-Line Breakdown

A spot crypto exchange built from scratch is about 304 engineer-weeks of work, which lands near $1.12m if you value an engineer-week at $3,700, and that figure covers software only: no licence, no liquidity, no market-maker retainer, no first year of operations. Crypto exchange development cost is really three budgets pretending to be one. There is the code, there is the permission to run it, and there is the money you have to put on your own order book before anyone else will. Most quotes you'll read online cover the first and skip the other two.

This article shows the arithmetic instead of the conclusion. Every rate and fee below comes from a source you can open, and every number I derived is labelled a model, so you can argue with the assumptions rather than the total.

The short version

• The scope, not the vendor, sets the price. A spot CEX is roughly 304 engineer-weeks; a hybrid build on a licensed core is 212; wiring up a white-label platform is 44.
• At $3,700 per engineer-week, derived from US Bureau of Labor Statistics wage data, that's about $1.12m, $784,000 and $163,000 in that order.
• White-label vendors charge a revenue share, not a build fee. HollaEx publishes 15 percent on its Enterprise Cloud plan, which flips the maths against renting somewhere past $6.4m of lifetime revenue.
• Licensing is the line nobody quotes. MiCA sets EUR 150,000 of permanent minimum capital for a trading platform, or a quarter of your fixed overheads if that's more. New York wants $5,000 to read your BitLicense application and a bond that starts at $500,000.
• Coinbase earned $4.06bn from transactions and $2.83bn from subscriptions and services in 2025. Its stablecoin line alone beat its entire institutional trading business by nearly three to one.
• Retail pays for everything. Coinbase's consumer take rate worked out at 1.39 percent of volume in 2025 against 4.9 basis points institutional, and institutions were 80 percent of the volume.

What crypto exchange development cost actually covers

An exchange does three jobs and everything else is decoration. It takes custody of other people's money, it matches buyers with sellers, and it settles the result correctly every single time. The trading screen is the part users see and the cheapest part to build.

Seven pieces carry the budget, and they aren't the seven a feature list would give you.

The matching engine keeps the order book and pairs orders by price-time priority in microseconds. Every bug here is a money bug. Match wrong and someone gets paid twice; lag under load and users get worse prices than the screen showed. Serious teams write it in a low-latency language, keep the book in memory, and treat the database as an append-only record of what already happened rather than the source of truth during matching.

Custody and key management is where exchange disasters start. Reserves split across an offline cold tier, a multi-signature warm tier and a small online float that pays withdrawals. How small? Coinbase is unusually specific in its 2025 annual report: it says it generally seeks to hold no more than 2 percent of assets under custody in hot wallets at any given time. Read that as one company's stated policy, not a sector benchmark, and be sceptical of any deck that quotes a precise industry-wide split.

Then there's the ledger, double-entry, reconciling to the chain and to your bank every day without a human touching it. That's the workstream teams forget to scope and then spend a quarter retrofitting.

KYC, AML and travel-rule plumbing. Identity checks at onboarding, sanctions screening, transaction monitoring that flags structuring and mixing, and audit logs you can hand a regulator without panic. Nobody enjoys building it. Skipping it kills exchanges.

Fiat rails mean card and bank deposits, payouts, chargeback handling and reconciliation. Banking relationships are slow to get and easy to lose, so the engineering is only half the problem here.

Risk and support tooling covers withdrawal limits, velocity checks, manual review queues, and an ops console a non-engineer can operate at 3am.

Liquidity isn't software at all, which is exactly why it gets left out of build quotes. More on that below, because it's the line that most often decides whether the other six were worth building.

If you want the component-level detail behind any of these, our web3 development guide covers the wider stack and blockchain security covers what goes wrong in production.

CEX, DEX or hybrid, and what each exchange model costs

The custody question decides your cost structure before you write a line of code. A centralized exchange holds funds and matches off-chain. A decentralized exchange settles on-chain through contracts and users keep their keys. A hybrid matches off-chain and settles on-chain.

A DEX moves the budget rather than shrinking it. You drop the custody and fiat workstreams and pick up a security bill you can't negotiate down, because the contracts are the product and they're public. If that's the direction you're leaning, our decentralized exchange development guide costs that build on its own terms, and DeFi as a category explains the liquidity mechanics a DEX inherits.

Most commercial projects still go centralized, because liquidity and fiat entry win customers. The rest of this article assumes a CEX unless it says otherwise.

What crypto exchange development cost looks like by route

Start with a defensible rate. The US Bureau of Labor Statistics puts the median annual wage for software developers at $135,980 as of May 2025. Wages aren't the whole cost of an employee: the same agency's Employer Costs for Employee Compensation release for June 2026 puts wages and salaries at 70.0 percent of total employer compensation in private industry, with benefits the other 30.0 percent. Gross the median up by that share and you get $194,257 a year, or $3,736 per engineer-week across 52 weeks. I round it to $3,700 below.

That's a US in-house number and deliberately conservative. Deliver through a nearshore team and your rate drops; hire senior matching-engine engineers in New York and it climbs. The engineer-weeks are the part worth arguing about. The rate is yours to substitute.

What drives crypto exchange development cost

This is a spot exchange: no margin, no derivatives, one fiat currency, ten to twenty pairs, web and mobile. It's a model based on how these builds have actually gone, not a quote, and the assumptions are the two paragraphs above.

Two things stand out. The trading interface is the third-largest line, which surprises people who think of the front end as the easy bit, and the four workstreams nobody demos (ledger, QA, infrastructure, admin) come to 100 of the 304 weeks.

The three routes, costed

White-label at 44 weeks isn't zero engineering. You still wire the KYC vendor, build the compliance policy behind it, integrate a bank, run a key ceremony, and build reporting the vendor's console doesn't give you. Roughly: 6 weeks of configuration and branding, 8 on KYC and AML wiring, 10 on fiat and banking, 6 on treasury procedures, 8 on reporting and support tooling, 6 on launch testing.

Hybrid at 212 weeks drops the matching engine and market data workstreams and adds core selection, forking and hardening. You keep custody, compliance and the clients, because those are the parts you can't afford to inherit from someone else's architecture.

Custom is the full 304.

The break-even nobody publishes

White-label looks like the obvious answer until you read the pricing model. Vendors charge on revenue share, not a licence fee. HollaEx publishes 15 percent on its Enterprise Cloud plan and no dollar figure at all, which is typical: the money follows your success rather than your budget.

Work it through. The gap between the white-label and custom figures is $962,000. At a 15 percent share, you hand that much to the vendor once cumulative revenue passes $6.4m. Assume all of it comes from trading at Coinbase's blended 2025 take rate of 0.33 percent (derived below), and $6.4m of revenue means roughly $1.9bn of traded volume.

Which means the real question isn't "which is cheaper". It's whether you believe you'll trade two billion dollars. If yes, build. If you genuinely don't know, rent, and negotiate the exit clause while you still have room to.

How long crypto exchange development takes, and who you need

Engineering is the shorter half of the schedule. That surprises people.

Divide the engineer-weeks by a real team and you get the calendar. A six-to-eight person squad is about right for this scope: two on the trading core, two on custody and settlement, two on clients, one on infrastructure, one on compliance integration, with QA shared. Push past ten and you spend the gain on coordination.

Now add the regulator, which runs on its own clock and doesn't care about your sprint board. Under Regulation (EU) 2023/1114, the competent authority acknowledges receipt within 5 working days, has 25 working days from your application to decide whether it's complete, then 40 working days from a complete file to grant or refuse, and 5 more to notify you. Add the windows that don't overlap and that's 70 working days of statutory maximum, call it 14 weeks, and the clock only starts when your file is finished. Building that file (governance, capital, AML programme, business continuity, legal opinions) is a months-long project of its own that runs in parallel with engineering if you're organised and after it if you aren't.

Then audits, penetration testing and a soft launch on a handful of pairs. A realistic total from kickoff to public launch on the custom route is twelve to eighteen months, so crypto exchange development cost reads better as a burn rate than a sticker price, and the variance sits almost entirely in the licensing and banking work rather than the code.

What blows a crypto exchange budget: liquidity, licensing and security

Exchange projects rarely fail on the matching engine. They fail on three lines that never appear in a software quote, and each can outgrow the crypto exchange development cost beside it.

Liquidity, and how new pairs get bootstrapped

An order book with no depth gives terrible fills, and traders leave for a venue where they can execute. New exchanges hit the cold-start problem head-on: traders want liquidity, and liquidity follows traders.

Four mechanisms actually get used, usually together.

• Pay makers to show up. Look at what venues charge and the strategy is written into the fee schedule. Kraken's published Pro spot fees start at 0.40 percent maker and 0.80 percent taker at the bottom tier and fall to 0.00 percent maker and 0.05 percent taker above $500m of 30-day volume. Binance's schedule runs 0.100 percent both sides for a regular user down to 0.011 percent maker and 0.023 percent taker at VIP 9. Zero maker fees aren't generosity. That's what a venue will pay to keep its book deep.
• Contract a market maker. A dedicated firm quotes both sides on your named pairs to an agreed spread and uptime, usually for a monthly retainer, a fee rebate, a loan of inventory, or some combination of the three. Nobody publishes these terms, so assume the first quote you get isn't the market.
• Route to someone else's book. Aggregators and partner venues fill orders your own book can't. Your users see depth; you see a thinner margin per trade and a dependency you now have to monitor.
• Run your own book, carefully. Seeding inventory and quoting yourself is the fastest way to look liquid on day one, and the fastest way to lose money if the strategy is naive. It's also the option a regulator will ask the most questions about.

Whichever you choose, liquidity is an operating cost that recurs every month, not a launch task. Underestimate it and you'll sink the launch faster than any technical failure will.

Security, and what an audit really costs

A CEX and a DEX spend security money in different places. The DEX pays for contract review, and those prices are public: Code4rena's audit listings for the year to September 2026 show awards from $4,000 for a small mitigation review up to $500,000 for the Monad audit, with most engagements between $20,000 and $110,000. Sherlock publishes a scoping guide that maps size to review time: about 3 days for 500 Solidity lines, 12 days for 2,000, 38 days for 6,000. Useful anchors even if you never run a contest, because they price the review of a given surface area.

A centralized exchange has far less on-chain code, so those numbers shrink. What replaces them is penetration testing of the withdrawal path, an independent review of key management, and a standing bug bounty. None of that is published, which is worth knowing when a vendor quotes you a flat security line. Our guide to smart contract audits covers what a review does and doesn't catch.

One audit isn't enough, either. Audit the deposit and withdrawal flow, the custody procedures and the contracts separately, and re-audit after any change to the money path.

Compliance, priced per user

Identity verification is one of the few compliance costs with a published price list. Sumsub charges $1.35 per verification on its Basic plan and $1.85 on the Compliance plan that adds AML screening, ongoing monitoring and proof-of-address checks, with monthly minimums of $149 and $299 and billing only on successful verifications.

Model it: onboard 50,000 verified users in year one on the Compliance plan and you spend $92,500. That's about 8 percent of the custom build, for one vendor, in one year, and it scales with growth rather than with scope. Transaction monitoring, sanctions screening and travel-rule messaging sit on top and are priced per firm rather than per user.

How crypto exchanges make money

Trading fees pay the bills. The assumption that they are the business doesn't survive contact with a real income statement.

Coinbase reported $7.18bn of total revenue for 2025. Transaction revenue was $4.06bn of it, and subscription and services revenue $2.83bn, with the rest corporate interest. Inside the second number, stablecoin revenue was $1.35bn, blockchain rewards (mostly staking) $677m, interest and finance fee income $247m, and other subscription lines $555m.

The stablecoin figure is worth a second look. It's nearly three times the $480m Coinbase earned from all institutional trading, and it comes from a share of the reserve income on USDC rather than from anyone placing an order. The 10-K also notes that one counterparty accounted for 19 percent of total revenue in 2025. Diversification, in this case, means a different concentration.

A venue big enough to earn $4bn from trading still had to build a second business to stay profitable.

The take rate, and who actually pays it

The same filing publishes trading volume, which lets you compute what a venue earns per dollar traded. Coinbase's 2025 volumes were $239bn consumer and $982bn institutional, $1,221bn in total.

Institutions were 80 percent of the volume and 12 percent of the transaction revenue. Retail funds the exchange, and every published fee schedule is built around that fact: high headline rates at the bottom tier, near-zero at the top. If your plan is to win institutional flow, budget for a business that earns five basis points and needs enormous scale to matter.

The other lines

Spreads. When an exchange makes its own market it keeps the difference between bid and ask. On thin pairs that spread can dwarf the headline fee, which is how some low-fee venues make their margin.

Deposit and withdrawal fees are the quiet ones. Crypto deposits are usually free as a hook, while fiat and on-chain withdrawals are where the meter runs, because the exchange is covering real network and banking cost.

Listing fees. Projects pay for distribution and credibility on a high-traffic venue. Lucrative and reputationally dangerous at the same time: list garbage and you'll spend the fee twice winning your users back. For the team on the other side of that fee, a listing is often the last step after they create a cryptocurrency.

Earn, staking and lending products keep a margin between what users receive and what the underlying protocol or borrower pays. Coinbase's $677m of blockchain rewards is what that looks like at scale. It's also the product line that draws the most regulatory attention, so the compliance work goes before the marketing.

Worth noting what none of this covers: the cost of running the thing. Coinbase spent $5.75bn in operating expenses against that $7.18bn of revenue, and technology and development was $1.67bn of it, under 30 percent. General and administrative ran $1.62bn, sales and marketing $1.06bn, transaction expense $1.02bn. Read the proportions rather than the magnitude. At the largest listed crypto exchange, building the software is a minority of what it costs to keep the lights on.

What licensing adds to crypto exchange development cost

Compliance stopped being a footnote and became the thing that sets your launch date. Unlike almost everything else in this article, parts of it have a published price.

The EU under MiCA

Operating a trading platform makes you a class 3 crypto-asset service provider. Annex IV of Regulation (EU) 2023/1114 sets permanent minimum capital by class: EUR 50,000 for a class 1 firm that executes, places, transmits orders or advises; EUR 125,000 once you add custody or exchange of crypto for funds; EUR 150,000 for class 3, which is what most people mean when they say exchange.

Don't read those as the cost of a licence. They're capital you hold, and Article 67(1) requires prudential safeguards equal to the higher of that floor or a quarter of the previous year's fixed overheads. A worked example: burn EUR 2.4m a year on fixed costs and the binding number is EUR 600,000, four times the headline. The requirement grows with the business rather than sitting still, which catches teams that budgeted once and moved on.

The application fee is separate and modest. Estonia's Finantsinspektsioon publishes a EUR 3,000 processing fee for a CASP authorisation, and confirms the same 25 and 40 working-day assessment windows MiCA sets out. What actually drains the budget is the file behind the application: legal opinions, compliance hires, monitoring vendors, audits. None of those publish a price list, which is why every cost guide skips them.

The United States

Two layers, and both bite. Federally, FinCEN requires MSB registration within 180 days of the business being established, renewed every 24 months by 31 December. There's no published filing fee, and registration is the easy part.

The states are where the money goes. New York charges a $5,000 application fee for a BitLicense, and the DFS rules set a surety bond or trust account that generally starts at $500,000 and rises with the business model. Multiply something like that across the states you want to serve and the licensing budget stops looking like a rounding error.

What is still unsettled

Stablecoins got a federal statute: the GENIUS Act became Public Law 119-27 on 18 July 2025. Market structure did not. The Digital Asset Market Clarity Act, H.R. 3633, passed the House on 17 July 2025 and has not been enacted, so spot trading in the US still runs on a patchwork of existing securities and commodities law rather than a purpose-built regime. If your business model depends on how a token gets classified, that classification isn't settled, and a plan that assumes otherwise is a bet rather than a budget.

Treat legal counsel as a first-week hire. It's the cheapest way to avoid discovering in month nine that your architecture assumed a permission you can't get.

Build, buy or white-label a crypto exchange

Six situations cover most of the real decisions.

And the pre-launch checklist, in the order that keeps you out of trouble.

Skip any row and you'll meet it again after launch, at a worse price.

Costing your own crypto exchange build

Crypto exchange development cost is a scope question before it's a price. Count the workstreams, put a rate on an engineer-week, and you get a number you can defend in a board meeting. Then add the two budgets the software quote left out: the capital and fees your regulator requires, and the liquidity you fund every month whether anyone trades or not. In the model above that's $1.12m of engineering for a custom spot CEX, $784,000 hybrid, $163,000 to wire up a white-label, and a licensing line that starts at EUR 150,000 of held capital in the EU before a single lawyer sends an invoice.

We've built this shape of product more than once. EVERSE is a centralized exchange trading over 150 cryptocurrencies on web and mobile, built from scratch with its own matching engine. Planetcoin took the opposite brief: get a first-time buyer from a debit card to holding crypto in seconds, with the fee visible before the buy is confirmed. All Crypto Mechanics folded margin and derivatives trading, hardware-wallet custody, a debit card and fiat ramps into one account, and Glue traded crypto against gold as the base asset with a supply-and-borrow lending side.

Choosing a partner rather than a route? Our rundown of blockchain development companies is a fair start, and the digital wallet types matter for custody before you commit to an architecture.