Equinix is expanding its BG2 data center in Bogotá as demand for artificial intelligence and cloud infrastructure grows across Latin America. DPL News reported on August 21, 2026 that the company invested an additional $28 million in the expansion, which adds 550 cabinets and 3 MW of electrical capacity. The original BG2 construction involved about $45 million and began operating in 2024.
The expansion is significant because Equinix is not simply adding conventional colocation space. BG2 is being prepared for higher density workloads, including GPU based AI infrastructure. DPL News reports that the facility uses liquid cooling to move heat away from high output components and that the site has been designed with a possible additional phase that could take electrical capacity to as much as 9 MW without expanding the physical footprint of the building.
That combination of more power in the same footprint captures the operational challenge created by AI. Sensaka's guide to liquid cooling in data centers explains why higher rack density is pushing operators toward cooling designs that can remove heat more effectively than conventional air systems alone. The Bogotá expansion shows that this is already becoming part of mainstream facility planning in growth markets.
AI changes the value of a cabinet
Traditional colocation growth could often be described through cabinet counts and floor space. AI makes those measures less informative unless they are paired with power density and cooling capability. Two cabinets can occupy the same physical area while supporting very different computing loads.
A rack full of conventional enterprise servers may operate within an established air cooled design. A rack containing modern accelerators can require substantially more power and produce far more heat. That means operators can no longer treat every available cabinet position as interchangeable capacity. The useful question is how much high density compute the electrical and thermal systems can sustain.
Equinix's decision to add capacity in Bogotá reflects this transition. The company is preparing the facility for customers that may need dense compute, private interconnection and access to cloud ecosystems at the same time. The commercial product is therefore becoming a combination of powered space, cooling capability and connectivity rather than simple square meters.
Bogotá is also an interconnection story
DPL News reports that Equinix's two Bogotá facilities connect with more than 40 network operators and more than 100 fiber routes reaching the sites. The company also points to private connectivity toward Miami and Atlanta as part of the broader role Bogotá plays in moving data between South America and North America.
This matters for AI because infrastructure performance depends on more than the accelerator itself. Training, inference and enterprise AI workflows can move large datasets between users, storage systems, clouds and model services. A data center that combines dense compute with a strong interconnection ecosystem can therefore provide a different kind of value from an isolated facility that has abundant power but weaker network reach.
The trend also supports distributed AI architectures. Not every workload needs to be concentrated in a single North American hyperscale campus. Inference and data sensitive workloads may benefit from being closer to regional users and enterprise data, especially when latency, regulatory requirements or data movement costs matter.
Sustainability claims will face more operational scrutiny
Equinix says BG2 operates with renewable electricity, primarily hydropower under supply contracts, and that the Bogotá sites also use smaller solar installations for common building services. Those measures sit within a larger industry challenge. AI demand is increasing total electricity consumption at the same time operators are under pressure to improve efficiency and reduce emissions.
The useful measure is therefore not whether a site can describe itself as efficient or renewable in general terms. Operators increasingly need to show how energy sourcing, cooling design, utilization and expansion interact as computing density rises. A facility can become more efficient per unit of compute while still consuming more total electricity because the amount of compute grows faster.
BG2 is a relatively compact example of that future. More capacity is being pushed into an existing location, with liquid cooling and interconnection becoming part of the design response. As Latin American demand grows, the competitive data center will increasingly be the one that can add dense capacity without losing control of power, heat and network performance.
Originally published on the Sensaka blog.
