Why we start with Hyperliquid
When I tell people our first venue is a decentralized perpetuals exchange, most of them expect to hear the name of a large centralized exchange. The choice comes down to what Hyperliquid publishes, and what that lets us build and prove.
I spent years working next to trading desks, and two complaints about market data came up again and again. The first was that nobody could check it. A vendor sends you a file of trades and book updates, and if something is missing or reordered you find out months later when a backtest refuses to match production. The second was that the most useful information was never in the file at all. You could see that 40 BTC traded at a price. You could not see whether it was one participant or forty, or whether the same address had been quietly building the position all afternoon.
Hyperliquid addresses both, and that is why it is the first venue we will ship.
The order book is on-chain
On a centralized exchange, the matching engine knows who placed every order, and that knowledge stays inside the exchange. The public feed shows aggregated levels, sometimes individual orders, and anonymous trades.
Hyperliquid runs its order book on its own chain. Every order, cancel and fill is a transaction, and every transaction carries the address that sent it. A node that follows the chain sees each order event with its block, its position in the block and its owner. That is what we call L4: order-level data with participant attribution. On most venues it does not exist outside the exchange. On Hyperliquid it is public by construction.
L4 changes the questions you can ask. You can measure how a single market maker manages inventory, how often the largest takers trade ahead of funding settlements, or what share of the displayed book comes from a handful of addresses. Our participant flow field starts from exactly this data.
Anyone can check our work
Because the source is a public chain, any customer can run their own node and compare our data against it, block by block. We think that is a healthy position for a data company to be in. It forces us to get sequencing, gaps and corrections right, because a mistake can be found by anyone who looks.
It also gives us a clean reference for our own validation. When we reconstruct order queues, we can check every fill against what the chain says happened, and we do. That work will be written up in detail once we have a full month of reconstructed books to report on.
It is a hard case, on purpose
Hyperliquid is not the easiest venue to capture. Timestamps are block times, so many messages share one value and ordering has to come from sequence. There are two sources to reconcile: the node's own output and the public WebSocket feed. Message rates during volatile hours are high, and order-level data multiplies them.
We wanted to start with the hard case. A pipeline that can capture, merge and normalize Hyperliquid at L3 and L4 from two regions has already solved the problems that matter on the centralized venues that follow. Adding the venues planned for Phase 2 is then mostly connector work and venue-specific rules, rather than a redesign.
It is a real market
None of this would matter if nobody traded there. Hyperliquid's BTC, ETH and SOL perpetuals carry billions of dollars of open interest, with books deep enough that execution and queue questions are worth asking. Funding settles every hour instead of every eight, which makes it a useful test for the cross-venue funding alignment we will need later.
What it does not give us
Being honest about the limits matters as much as the reasons. An address is not an identity. One firm can trade from dozens of addresses, and one address can run several unrelated strategies, so L4 on its own tells you which address acted and leaves open who stands behind it. Linking addresses to each other and to activity on other venues is a separate problem, and it is what our chain intelligence work is for.
Block timestamps also put a floor on timing precision. Everything in one block shares a time, so questions about which of two orders in the same block came first have to be answered from sequence, not from the clock. And running our own nodes in two regions costs more than subscribing to a WebSocket feed. We think the order-level data is worth it.
What we have so far
We have recorded Hyperliquid internally since November 2025, from Tokyo and from Ireland, with every message stamped on a PTP-disciplined clock and the two regions merged by venue sequence number. That archive is what our research runs on today, and we plan to publish the first notes from it over the summer. It is not for sale yet. Our target is to open early access in Q1 2027.
Coverage for Hyperliquid at launch is perpetuals, funding, full-depth L2, L3 and L4, plus the extended fields computed from them. The details are on the Hyperliquid coverage page.
What comes next
Hyperliquid on its own does not describe the crypto market, and we have never pretended otherwise. Binance, Bybit and OKX are planned for Phase 2 in 2027, then Coinbase, Deribit, Kraken and Bitget for Phase 3 in 2028, each subject to a data agreement with the exchange. The roadmap has the full sequence, and we will move venues up the list where enough partners ask for them.
If you trade on Hyperliquid, or research it, I would like to hear what you need from the data. My address is daniel@financedatacloud.com, and I read everything that comes in.
Order flow with names attached.
Pre-register for early access to Hyperliquid trades, full depth, funding and L4 order flow, targeted for Q1 2027.
EARLY ACCESS · PRE-SALE PRICINGPre-register for early access → DANIEL GARCIA · CEO & FOUNDERdaniel@financedatacloud.com →