For most of the last decade, sub-millisecond execution was a private club. Colocation cages inside NY4, LD4 and TY3 belonged to a small set of prop shops and tier-one banks — and the rest of the market paid a quiet, structural tax on every fill.
That equation is changing. Cloud-adjacent matching engines, private fibre between venues, and lightweight FPGA edge nodes are collapsing the geographic advantage. The question is no longer whether you can be fast; it is whether your routing logic can decide, in the same window, where fast actually pays.
The new colocation grid
Our routing fabric now touches nine points of presence across four continents. Every quote enters a normalisation layer within 90 microseconds of arrival, and the ensemble decides in under three milliseconds which venue will hold the fill without slippage.
- Zurich (ZRH-01) — European majors, FINMA-supervised custody link
- London (LD4) — cross-venue FX aggregation
- New York (NY4) — equity + index derivatives
- Tokyo (TY3) — Asia session bridge
Why speed alone is not the story
Every additional millisecond you shave off the wire is only worth what the venue on the other side will honour. The real work is not the transport — it is the pre-trade decision that decides which venue is currently truthful about its book.
"Latency without context is just heat. What matters is a router that knows when the fastest path is also the wrong one."— A. Takeda, CTO