BlackForge vs Kaiko (and Amberdata)
Updated
BlackForge and Kaiko sell the same class of data, order-book microstructure, and reach the buyer in two completely different ways. Kaiko, which acquired Amberdata in 2026, is the institutional counterparty: history back to the middle of the last decade, regulated benchmarks, and a landed contract in the five figures. BlackForge is the credit-card subscription: computed columns across nine venues and roughly 11,800 pairs, self-serve, starting at $29 a month. The whole comparison comes down to which of those two buyers you are.
What is a cheaper alternative to Kaiko and Amberdata for order-book data?
Kaiko, which now owns Amberdata, sells institutional order-book feeds from five-figure annual contracts, with history to 2015 and regulated benchmarks. BlackForge sells computed order-book microstructure across nine venues and roughly 11,800 pairs, self-serve from $29 a month. Same class of data: one is an enterprise contract, the other is a credit-card subscription.
One company, two SKUs
Kaiko and Amberdata are the same company. The acquisition was announced on 2026-06-02, and the two products still sell under two names with two separate checkouts, so it is easy to weigh them as rivals when they are one vendor. We treat them as one here, because whichever door you walk through, you are buying from the same house.
The reason that matters for pricing is that Amberdata publishes a self-serve plan and Kaiko does not. Amberdata's checkout is $600 a month or $5,000 a year per exchange-market, across 51 exchanges (intelligence.amberdata.com/plans, read 2026-08-12). Nine venues, the set BlackForge collects, therefore land at $5,400 a month or $45,000 a year, and that buys the raw order books with no computed metrics layered on top. That $5,400 figure is the anchor for everything below.
Kaiko's side is quoted higher and sold through sales. Its L2 order-book tick product starts at $2,500 a month, and the average landed contract is a five-figure annual number (kaiko.com/products/l1-l2-data, read 2026-08-12). Neither of those is a bad price for what an institution gets. They are simply prices for a different buyer than the one who wants to put a card in and start querying this afternoon.
| Kaiko / Amberdata | BlackForge | |
|---|---|---|
| How you buy | Sales cycle; Amberdata has a self-serve tier | Self-serve, card, no call |
| Entry price | Amberdata $600/mo per exchange-market; Kaiko L2 from $2,500/mo | Pro $29/mo; Max $99/mo; Ultra $399/mo |
| Nine venues of books | $5,400/mo or $45,000/yr on Amberdata, raw | Included; every venue, nothing gated |
| Computed microstructure | Depth-at-N%, spread, slippage, VWAP, indices | 120 columns per pair per window |
| History | Back to 2015 (Kaiko) and 2017 (Amberdata) | From mid-2026; retention infinite going forward |
| Resolution | Tick-level | 5-minute window |
What $5,400 a month buys, and what it does not
On Amberdata, the nine-venue number gets you the raw order books: the levels, the updates, the tape. That is the complete raw feed, and for a desk that wants to run its own models it is exactly the right shape to buy. What it does not include is the computed behavioural layer, the columns that describe how the book moved rather than what it looked like at one instant.
It is worth being precise and fair here, because both vendors ship real analytics and it would be wrong to call them raw-only. Kaiko and Amberdata both publish a derived layer: depth-at-N-percent, spread, slippage, VWAP, and a catalog of indices and reference rates. What neither ships as columns is the behavioural block: level flicker, resting-level lifetime, gross liquidity removed from the book, and aggressor-bucketed counts of large trades. Those measure how the book churned within the window, and they are the columns you would otherwise write yourself from raw ticks.
BlackForge computes that block for every pair on all nine venues at five-minute resolution, with nothing scoped out by liquidity. The full universe is the product, so a thin pair on a smaller venue gets the same columns as BTC on binance. Here are three of those behavioural columns, measured on the largest venue in the most recent closed window.
BTCUSDT on binance
- Bid liquidity removed
- $283.9M
- Median level lifetime
- 1,800ms
- Level flicker count
- 7,970
Priced side by side
The cleanest way to see the gap is to put the monthly cost of nine venues of order-book data on one axis. Raw on Amberdata, a tick floor on Kaiko, computed on BlackForge Max. These are the same unit, dollars per month, so the bars mean the same thing.
| Label | Value |
|---|---|
| Amberdata (9 venues, raw) | 5,400 |
| Kaiko L2 tick (floor) | 2,500 |
| BlackForge Max (computed) | 99 |
BlackForge Max is $99 a month, self-serve, for all 120 columns. Ultra is $399 a month for the full archive and scale. The point is not that BlackForge is a replacement for a Kaiko contract line by line, it is not, but that a large fraction of the people quoted $45,000 a year for raw nine-venue books wanted computed microstructure they could query today, and that is a different product at a different price.
Breadth across venues, nothing scoped out
Every venue BlackForge collects gets the same computed columns, and every pair on it. There is no majors-only tier and no per-exchange-market line item. The depth column below is one of the 120, read live across four venues in the same window, so you can see the same measurement land on each.
| Label | Value |
|---|---|
| binance | 48,054,761.578 |
| coinbase (BTC-USD) | 13,559,124.654 |
| kraken | 2,878,718.853 |
| okx | 3,669,493.6 |
| Kaiko / Amberdata | BlackForge | |
|---|---|---|
| Account to start | Sales contact; Amberdata has self-serve checkout | Card, no call |
| Minimum commitment | Five-figure annual on Kaiko; per exchange-market on Amberdata | Monthly, cancel anytime |
| Computed behavioural columns | Not shipped as columns | 120 columns per pair per window |
| History depth | Back to 2015 / 2017 | From mid-2026 forward |
Where Kaiko and Amberdata wins
Kaiko and Amberdata
- History that goes back to 2015 (Kaiko) and 2017 (Amberdata); BlackForge collection started in mid-2026, so a decade of in-sample depth is theirs and not ours.
- Coverage far beyond nine venues, more than a hundred on Kaiko and 51 on Amberdata, plus derivatives markets BlackForge does not touch.
- Regulated indices and reference rates, the benchmark products an institution needs for pricing and reporting.
- The vendor-risk-review credibility of an established company, uptime history and a large client base, which a small operator cannot match in a procurement review.
Choose them if you need a decade of history in-sample, a procurement-approved institutional counterparty, regulated benchmarks, or coverage far beyond nine venues.
Are Kaiko and Amberdata the same company?
Yes. Kaiko acquired Amberdata, announced on 2026-06-02, so they are one company. They still sell under two names with two separate checkouts, which is why they are easy to mistake for competitors. Amberdata has a public self-serve plan and Kaiko is sold through a sales cycle, but the money goes to the same house either way.
How much is Amberdata for nine venues?
Amberdata's checkout is $600 a month or $5,000 a year per exchange-market, across 51 exchanges (intelligence.amberdata.com/plans, read 2026-08-12). The nine venues BlackForge collects therefore come to $5,400 a month or $45,000 a year, and that price is for the raw order books with no computed metrics layered on. BlackForge Max is $99 a month, self-serve, for all 120 computed columns.
Does Kaiko compute the same metrics as BlackForge?
Partly. Kaiko and Amberdata both ship a derived layer: depth-at-N-percent, spread, slippage, VWAP, and a catalog of indices and reference rates. What neither ships as columns is the behavioural block BlackForge computes: level flicker, resting-level lifetime, gross liquidity removed from the book, and aggressor-bucketed counts of large trades. Those describe how the book churned, not just how it looked at one instant.
Can I get institutional history from BlackForge?
No. BlackForge started collecting in mid-2026, so there is no decade of back history to sample, and retention is only infinite going forward from that start. If you need order-book history reaching back to 2015 or 2017, that is a structural advantage of Kaiko and Amberdata, and it is the main reason to choose them over BlackForge.
See the full metric reference, or browse all comparisons.