Alberto Caraballo
A network engineer planning a route into Mexico five years ago made four separate decisions. Which submarine cable to land on. Which backhaul provider would carry the traffic inland. Which border facility to cross at. Which U.S. interconnection point to reach on the other side. Four decisions, four sets of counterparts, four commercial conversations that rarely spoke to each other.
The individual sites weren’t the hard part. The connections between them were.
That design reality is starting to look outdated. The traffic moving through Mexico today treats the coast and the border as two ends of the same route, not two separate markets. The infrastructure is starting to catch up.
The traffic moved first
Mexico carries roughly 35 Tbps of international internet capacity, according to TeleGeography data compiled by LACNIC. That capacity has been growing at a pace consistent with the broader Latin American market, which TeleGeography projects at a 30% compound annual growth rate through 2030. The headline number matters less than what sits behind it: content providers, not traditional carriers, are now the fastest-growing source of that demand. On U.S.–Latin America routes, TeleGeography expects content networks to account for the outright majority of international bandwidth within the next year.
What does this look like on the ground? It looks like CDNs and cloud providers choosing interconnection points close to where end users actually live. It looks like AI inference workloads, which run continuously and serve users in real time, gravitating toward metro ecosystems with dense peering rather than remote sites with cheap power. It looks like a single flow starting at a landing station in Cancún, traveling inland through Querétaro, and exiting to the United States through McAllen or El Paso, all within one commercial envelope.
The pattern is visible at the border, where mixed peering communities bringing together global content providers and Mexican operators now handle volumes that would have been niche five years ago. It’s becoming the default shape of how Mexican traffic reaches the internet.
More points isn’t the same as better architecture
Mexico’s digital infrastructure has been getting denser for years. Querétaro alone is expected to host around 70 new data centers over the next three years. New submarine cables are landing on both coasts. Border markets that were secondary a decade ago, like El Paso, have matured into ecosystems of their own. Density, in raw terms, is not the problem.
The problem is that density on its own does not make a route. Adding another facility only helps the operator planning a path if that facility works under compatible rules with the other facilities on the route. Access terms, handoff procedures, and policy stability have to line up from one site to the next. When they do, every new location makes every existing location more valuable. When they don’t, the map keeps filling in while the friction between segments stays exactly where it was.
The same pattern shows up in how border crossings are maturing. When multiple crossings in different cities operate under the same access rules, an operator can plan redundancy across them as one commercial decision instead of two. The density is useful because the points are orchestrated to work together, not just co-located.
A different question for whoever designs the route
The practical question for operators planning international routes into Mexico is starting to change. It used to be about site selection: which landing, which crossing, which interior hub. Those decisions still matter, but they’ve become parts of a larger question about whether the full trajectory works end to end.
The recent MANTA landings in Cancún and Veracruz are a concrete test case. The system connects Mexico’s Gulf coast to Florida and branches toward Panama and Colombia, which means traffic entering Mexico from South America can now reach the U.S. through a route where the coastal landing and the border exit operate under the same commercial model, run by the same operator. That specific configuration didn’t exist in Mexico before. The interesting question isn’t whether it works for MANTA. It’s whether it generalizes, and whether operators start designing around the possibility.
Workloads that are growing fastest are the ones that care most about this. Content delivery architectures need predictable handoffs between international transit and domestic distribution. AI inference depends on consistent peering with local ISPs. Enterprise traffic that has to comply with both U.S. and Mexican data regulations can’t afford to renegotiate governance at every commercial boundary on the path. All of these are easier to design when the path answers to one logic instead of four.
The shift in one sentence
Mexico’s international networks used to be built as a collection of strong individual points. The capacity was there, but the route was something each operator had to stitch together themselves. What’s starting to emerge is different: infrastructure designed so that the points work together from the coast to the northern border, under rules that stay consistent along the way. The work of designing a route becomes simpler when the route is built to hold together.
