A cross connect is a direct physical cable between two parties inside the same data centre: your rack and a counterparty, an exchange, or a carrier. It bypasses the public internet and any shared network in between, so traffic crosses a single piece of fibre or copper rather than a series of hops.
Why it matters for latency
Every hop adds delay, and every shared link adds variance. A cross connect removes both: the path is dedicated and its length is fixed, so the latency is not only lower but predictable. In electronic trading the predictability often matters more than the absolute figure, because a route that is fast on average and occasionally slow is harder to work with than one that is consistently moderate.
How it works in practice
The data centre operator runs the cable through its meet-me room, the neutral space where tenants interconnect. You order it by specifying the two endpoints and the media type, and you pay a recurring fee per connection. Nothing routes: the two ends are physically joined.
Capacity is whatever the optics at each end support, commonly 10, 40 or 100Gbps for trading infrastructure. The constraint is rarely the cable.
Cross connect, peering and transit
These are often confused. Transit is buying access to the wider internet from a provider. Peering is exchanging traffic with another network, often across a shared fabric. A cross connect is just the physical link, and can carry any of those arrangements or a private conversation that is none of them.
In trading infrastructure
Cross connects are how a trading system reaches an exchange matching engine when both sit in the same facility, and they are the reason colocation is valuable in the first place. Colocation Hosting places your equipment in those facilities, and Venue Access covers connectivity to the venues themselves where a direct cross connect is not the right arrangement.
