The Americas need compute next door.

AI capacity is limited less by chips than by power, space and the time it takes to energize a site. In established US metros, that time keeps getting longer.

The border metros of the Americas have power, space and fiber within one hop of the US interconnection hubs. dense² turns that into reserved clusters US teams can buy the way they already buy: a fixed topology, a delivery date and named support.

Training and inference both want dense, dedicated clusters. New capacity in the largest US markets often waits on interconnection, construction and equipment lead times.

Teams that need GPUs this year end up negotiating quota, accepting best effort hosts, or waiting. None of those is a plan.

A border metro is not offshore in any sense that matters to a network. Typical round trips to the Central US stay under 20 ms, within the range of routes US traffic already takes every day.

Power, space and fiber sit where the capacity is needed, without waiting in a US metro queue.

A training run or a production inference fleet needs a fixed topology, a known date and a cost that does not move. On demand capacity offers none of the three.

We sell reserved clusters and GPU hours on 12 to 36 month terms. We do not sell tokens and we are not a marketplace.

Every site runs the same block: 256 NVIDIA B300 GPUs on 32 HGX nodes, one rail aligned fabric, managed Kubernetes on bare metal.

A standard block means procurement, operations and your engineers learn it once. The second site is a contract amendment, not a new project.

Our capacity sits in Mexico, and every document says so. That makes dense² a strong fit for startups, AI labs, neoclouds and cross border AI, and the wrong fit for US only hosting or FedRAMP.

We would rather lose a deal on the first call than on the compliance review.

Each border metro pairs with the US market across the line: Monterrey with Texas, Tijuana with California, Cancún with Florida. Then the wider region, block by block. The plan is 1,024 GPUs in the first 12 months: 256 in Q2 2027, 512 by Q4 2027 and 1,024 by Q2 2028.

dense² is being built to become the reference GPU capacity platform of the Americas. If that is the future you see too, we would like to talk.