4 Premium PoPs and Counting: Our 2026 Expansion Roadmap
We crossed 4 Premium Points of Presence this week. Here's a candid look at where we've been, where we're going, and the honest trade-offs involved in expanding a private backbone.
Where We Started
When RouteKey launched in Q3 2023, we had 1 PoP — which is where our seed customers were. Growing a private backbone isn't like spinning up cloud regions. Each PoP requires colocation contracts, cross-connects, transit agreements, and at least one peering relationship that's worth anything. Getting our first additional premium PoP by end of 2023 required extensive qualification of vendors.
By mid-2024 we had 3 Premium PoPs and had learned a hard lesson: raw PoP count is a vanity metric. What matters is the graph — specifically, how many useful paths exist between any two customer endpoints. A PoP in Lagos is useless if your only exit to the transatlantic backbone is via a degraded undersea segment.
Strategic Expansions
Our recent expansion added hand-picked routing locations heavily weighted toward strategic corridors that had been systematically underserved by enterprise-grade routing infrastructure.
The latency improvement for traffic traversing these new regions has been significant:
- London → Singapore: 155 ms → 112 ms average
- Singapore → Tokyo: 42 ms → 28 ms average
Across our exclusive customer base, this deliberate expansion allows for highly optimized dedicated flows.
What We're Building in H1 2026
Our next target locations are strictly vetted. The prioritization logic is straightforward: we model the expected latency improvement per dollar of infrastructure spend for each candidate location, weight it by the number of existing exclusive customers who would benefit, and rank accordingly. The current top of that list involves reinforcing our trans-pacific links.
The Honest Trade-offs
Expanding fast has costs. More PoPs means more operational complexity, more potential failure domains, and more surface area for our edge AI to manage. We've had two cases in the past quarter where a newly onboarded PoP had subtly incorrect routing advertisements that took 40 minutes to detect. Neither caused customer-facing impact, but they were humbling.
We've responded by tightening our PoP onboarding checklist — every new location now has to pass 72 hours of shadow-mode operation (receiving telemetry, not yet carrying live traffic) before going into production. It adds a week to each launch, and it's worth it.
Looking Ahead
We're not chasing a number. We're at 4 Premium PoPs today. Getting our next few locations right — especially in markets with less developed interconnect ecosystems — is the deliberate work of the next 18 months.
We'll update this post when we expand further. In the meantime, our system status is always current at /status.