Scalability
100,000
transactions per second across 25 distributed validators
~500 ms
median finality
Hathor’s consensus has reached 100,000 transactions per second with 25 validators distributed across US-East, US-West, UK, Europe, Tokyo and Sydney.
25 validators
distributed around the world
Global
US-East, US-West, UK, Europe, Tokyo and Sydney
Consensus designed around payments
Hathor’s new payments chain uses Byzantine fault tolerant (BFT) consensus built around independent payments. Validators confirm payments without a single leader setting the order for the whole network. A payment becomes final when two thirds of the committee approve it.
No single leader
Payment confirmation does not depend on one validator ordering every transaction. Validators can confirm independent payments as they arrive.
Predictable voting overhead
Consensus runs in regular voting rounds instead of adding a separate vote for every payment. In benchmarks, rounds stayed near half a second as payment volume increased.
Finality you can build on
A committee approval gives applications a clear settlement event, so they can act on a completed payment without waiting for successive block confirmations.
Distribution does not reduce TPS
Hathor kept throughput above 100,000 transactions per second even with 25 validators distributed worldwide. The added network latency increases finality time, while TPS remains at the same level.
Where the validators run
Network / setup
Transactions per second
Finality (p50)
One data center
inside the data center
102,888
not comparable
North America
N. Virginia, N. California, Montreal, Oregon
4 validators, public internet
98,244
404 ms
Global
US-East, US-West, UK, Europe, Tokyo, Sydney
25 validators, public internet
101,500
430–576 ms
SUB-SECOND FINALITY
A payment is final the moment the validators approve it. There are no confirmations to wait for, and nothing can undo it later. Adding validators or distance barely changes that time.
Adding validators barely slows settlement
4 validators
130 ms
10 validators
134 ms
16 validators
147 ms
25 validators
178 ms
Earlier benchmark: median time to final at 20,000 payments per second, in one data center. Volume held at 20,000 for every committee size, and six times the validators added only 48 ms.
430 ms
Lowest per-node p50 finality during the 600-second run with 110,000 transactions per second injected.
576 ms
Highest per-node p50 finality, recorded in Sydney. Across all 25 validators, each node’s p50 was between 430 and 576 ms.
Hathor verifies independent payments in parallel instead of forcing them through a shared account balance. Validator operators can increase verification capacity with more CPU cores or a GPU, while every validator still checks every payment.
Verification capacity on standard hardware
AMD EPYC, 16 cores
23,800
AMD EPYC, 32 cores
48,500
NVIDIA L4
101,700
NVIDIA RTX 4070 Ti
197,700
NVIDIA L40S
360,800
Verification benchmark with level 4 confidentiality and no consensus running. These are per-machine verification limits, not end-to-end network throughput. CPUs are 4th generation AMD EPYC servers on AWS.
Built for busy merchants
Each payment spends existing coins and creates new ones for the receiver. Payments to the same merchant can therefore be verified side by side, instead of competing to update one shared account balance.
Confidentiality at scale
High payment volume and confidentiality belong in the same experience. Earlier distributed tests reached 99,826 confidential transactions per second with amounts and assets hidden at level 4, while every validator checked every payment.