For most enterprise buyers, capacity purchasing is about throughput and price per Mbps. For a narrow but extremely demanding slice of our customer base — proprietary trading firms, exchanges, and market makers — the entire purchasing decision comes down to one number: deterministic round-trip latency, measured in fractions of a millisecond, on one specific route.
The physical floor on latency between two points is set by the speed of light in fiber (roughly 2/3 the speed of light in a vacuum, due to the refractive index of glass) and the physical cable route length — there’s no way to route data faster than light travels through the glass it’s traveling through. Our job on a latency-sensitive route is almost entirely about minimizing route length and minimizing the number and processing delay of intermediate regeneration and switching points, not about raw bandwidth.
On Meridian-1, for example, the New York–London round-trip latency floor set by cable length alone is close to 38.2ms. Our actual measured round-trip figure on a direct wavelength service is 38.5ms — within roughly 0.3ms of the physical minimum, the remaining difference coming from terminal equipment processing delay at each end rather than any additional route length. That 0.3ms gap is a real engineering target we track and periodically re-optimize against, because for a trading firm, a competitor shaving even a fraction of that gap represents a measurable execution-speed advantage.
This is also why physical route straightness matters disproportionately for this customer segment: a route that’s 200 km longer than the geometric minimum, even if it has more total capacity or a lower price per Mbps, is simply the wrong product for a latency-sensitive buyer — no amount of switching or amplification improvement compensates for extra physical distance.
