How to build a crypto exchange in 2026: the parts, the costs, the order
What an exchange actually consists of, what each part costs to build or buy, and the sequence that gets a licensed operator live without burning a year.
BitPolaris Engineering
The nine parts
Every exchange, whatever the brand, is the same nine systems. Underestimating any one of them is how projects slip from months to years.
| Part | What it does | Build in-house | Typical shortcut |
|---|---|---|---|
| Matching engine | Order book, price-time priority, order types | 4–8 months | Open-source engine (Go) |
| Core & ledger | Accounts, balances, deposits, withdrawals, fees | 6–12 months | Exchange core |
| Custody | Wallets, keys, chain scanners, policy | 6–12 months | Vendor custody or own layer |
| Liquidity | Someone must quote both sides of every market | Capital + MM desk | Mirror / bridge engine |
| Identity & AML | KYC tiers, screening, records | 2–3 months | Sumsub / Didit integration |
| Web & apps | Terminal, easy buy, iOS, Android | 6–9 months | Existing apps, rebranded |
| Operations console | Everything your team does daily | 6+ months, always underestimated | Rarely available off the shelf |
| Fiat rails | Bank in/out, PSP | 2–4 months per rail | PSP partner |
| Content & SEO | How users find you without ads | Ongoing | CMS + content licence |
Where the money goes
A credible in-house build is a team of eight to twelve for twelve to twenty-four months: roughly €1–2M before the first customer trades. The two line items people miss are the operations console, because nobody enjoys building it, and liquidity, because it is not software at all but capital. Filling your own books on a hundred markets needs close to a million dollars of inventory sitting idle.
The order that works
- Licence path first. The regime decides where keys live, which KYC vendors you may use, and whether you can host in the cloud at all.
- Custody and ledger before anything users see. If this layer is wrong, nothing above it can be trusted.
- Liquidity strategy before listing markets. Decide: own capital, external market makers, or a mirroring engine that hedges each fill.
- Operations console before launch, not after. Your compliance team will refuse to go live without approval queues and an audit log.
- Apps last, but start store review early; for crypto apps it is the least predictable step.
Build, buy, or license
Build if exchange technology is your edge and you have the team. Buy a white-label if you want speed and accept the vendor holding your keys and your uptime. License production software if you want speed and control: the code already runs an exchange, you can self-host it, and the operations console exists. The honest trade-off is that you inherit someone else’s architecture decisions; ask to see them running before you sign.
BitPolaris licenses the production software behind a live exchange: self-hosted, with the operations console and a liquidity engine that needs no inventory.