Kaiko vs Amberdata

The acquisition changes nothing on the shelf yet

Kaiko announced that it had acquired Amberdata on 2 June 2026. Four months later, the purchase in front of a buyer is still two purchases. Both cards record separate sites, separate documentation hubs on separate hosts, separate keys and separate contracts. Re-read on 3 October 2026: the announcement, in a quote from Amberdata's chief executive, speaks of consolidating "resources and solution sets" and says nothing about which product line, brand or contract survives; Amberdata's pricing page still sends buyers to its own store, and its changelog carries no entry about Kaiko at all — the most recent one, posted 17 August, is the notice of BitMEX's closure on 23 September.

So the honest answer to "which one, now that they are one company" is: whichever has the data you need today, on the shortest commitment that covers it. Nothing published says what either contract turns into, and a buyer cannot price a merger that has not been described.

How you can buy each one

Amberdata sells one exchange and one market at a time, with a card. 600 dollars a month or 5,000 a year per combination, from 51 listed when the card was checked in September — Binance spot, Deribit options, Hyperliquid futures. The monthly plan carries one year of history; the annual one carries full history and S3 delivery. Keys arrive in 24 to 48 business hours, and a call that does not name the exchange is rejected. Bulk delivery across everything you have bought is a separate line the store does not draw, recorded on the card at 19,580 dollars a year.

Kaiko sells nothing without a conversation. No price, no self-serve key, a trial of unpublished length. The quote is built on instruments, data type, granularity, live against historical, and usage, and almost every dataset in its dictionary names the tier it belongs to.

That asymmetry is the first fork. A reader who needs one venue's tape this month has one option. A reader whose requirement spans dozens of venues is buying an enterprise contract from either vendor — Amberdata's enterprise tier is quote-only too — and should be comparing quotes, not list prices.

Order-book history flips with the delivery channel

This is the comparison that neither card makes alone, and the one most likely to decide a research budget.

Through an API, Amberdata is deeper. Its REST order-book snapshot and event endpoints return the past 18 months on every exchange — a cap introduced in May 2025. Kaiko's REST carries no full order book at all, and its spread, depth and slippage metrics sit in a one-month rolling window; its gRPC stream replays 72 hours, which is a reconnection buffer and not an archive.

In bulk files, Kaiko is deeper. Its raw order-book snapshots — at least one a minute, to 10% of depth — go back to 2015 in cloud delivery, varying by exchange, with a full-book CSV from August 2023 and top of book from December 2022. Amberdata's earliest order book on any venue is January 2018, on Kraken, and anything older than 18 months comes by S3 or Snowflake.

So a desk that wants to page a year of books out of an API in code is an Amberdata customer, and a desk that wants the 2017 cycle's books in a warehouse is a Kaiko one. Trade history leans the same way as the archive: tick trades to 2010 on Kaiko against January 2013 on Amberdata, with Amberdata's OHLCV reaching 2011 on Bitstamp.

Rate limits, in the same unit

Kaiko's standard REST limit is 6,000 requests a minute per key — 100 a second — with no daily ceiling stated. Amberdata's self-serve plan allows 20 a second and 250,000 a day, and the second number is the one that binds: at full speed the day's allowance is gone in about three and a half hours, which averages to under three requests a second across a day. Enterprise keys run at 60 a second. Amberdata also rejects any REST result over 10 MB and times out at 28 seconds, which is the documentation telling you that bulk work belongs in the files.

Where each one is the only answer

Derivatives structure is Amberdata's. Volatility surfaces by delta and by moneyness, term structure, implied against realised, variance risk premium, funding rates on 16 of 20 futures venues reaching back to May 2016. Kaiko's implied-volatility surfaces start in April 2021 and its liquidation events only on 8 January 2025. Amberdata's liquidations cover ten of its twenty futures venues, so an aggregate built on them describes those ten rather than the market.

Benchmarks and deep tick history are Kaiko's. It describes its indices arm as a BMR-registered benchmark administrator, which matters if a fair value has to survive an auditor, and it is the one of the two with trades before 2013. The indices are a separate purchase from the market data.

Solana is neither's. Kaiko's standard package carries 30 days of rolling Solana history, with inception-to-date as a custom deal. Amberdata lists Solana among chains "available with select endpoints, as backfills and data verification are ongoing", and has no Base at all. If deep Solana on-chain history is the requirement, neither standard offer is it.

Which one, and when

Take Amberdata if you need one or a few venues' feeds now, if the job is options and volatility analytics, or if you want order-book history through an API rather than a warehouse. Buy the monthly plan if a year of lookback is enough, because with the merger undescribed the shortest commitment is worth something; take the annual one the moment you need full history.

Take Kaiko if you need trades before 2013, order books from the years before 2018, a regulated fair value, or cross-venue volume that a 250,000-request day cannot carry. Expect a sales cycle, and ask in it what happens to the contract if the two product lines are combined — nothing on either site answers that.

Do not buy both expecting them to be one. Two keys, two hosts, two schemas, two invoices. If the plan is to join them, budget the reconciliation as your own work, because neither vendor has published a combined product to do it for you.