Amid all the prompt-tweaking and chatbot building, it’s easy to forget about the physical aspect of AI. Ultimately, these are machines that run on power, water, land, and a whole lot of infrastructure most people never see.
Carolina Muñoz López has spent her career thinking about infrastructure in exactly those terms. Trained as a civil engineer in Colombia and now working at Coral Gables investment firm Tallvine Partners, she looks at AI not as software, but as systems that include energy, cooling, land, and everything in between.
“AI has different requirements,” Muñoz López told Refresh Miami. “It’s not just about the technology. It’s about power, cooling, and how you connect all of that to the user.”
Not all AI infrastructure is created equal. The massive data centers used to train AI models are unlikely to land in South Florida. Those tend to cluster in places like Northern Virginia and Texas, where energy is cheaper and risks are lower.
Miami, instead, is playing a different role.
“Miami will become important not for the training of AI, but for the latency,” she said. “The connection with the end user is what matters here.”
That “last mile” layer – where data is processed closer to users – is where Miami has a real shot. It requires less space and power, but it still brings challenges.
The biggest shift comes down to density. AI workloads require far more power per rack than traditional systems. “The demand increases six to eight times,” Muñoz López said.
More power means more heat. And that brings the conversation to cooling.
“When you pack much more power into the same space, you generate an extraordinary amount of heat,” she said. “Cooling is one of the most underappreciated challenges right now.”
Traditional air systems are starting to fall short. The industry is moving toward liquid cooling, circulating fluid through hardware to remove heat more efficiently.
In Miami, the heat and humidity make cooling systems work harder, increasing energy use. Water also becomes a constraint, pushing operators to explore recycling systems.
“It’s not about land anymore,” she said. “It’s about how you cool.”
“The big problem is not today,” Muñoz López explained. “It’s in the next ten years.” Florida has enough supply to support near-term growth in smaller data centers. But that may change.
That’s where the concept of interconnection comes in: getting energy from where it is produced to where it is needed. “You can produce energy,” she said. “But how do you move that to the consumer?”
Grid infrastructure takes years to build, while new energy sources can come online much faster. The gap between the two is growing.
All of this puts Miami in an interesting position. It may not become a hub for training AI models, but it does not need to.
Rather than competing with the largest markets, she believes Miami should focus on proximity, connectivity, and speed. “Don’t try to compete with Texas or Virginia,” Muñoz López said. “Focus on where you have an advantage.”
“Miami is the gateway to Latin America,” Muñoz López continued. “We already have strong fiber interconnections. Let’s leverage that.”
READ MORE IN REFRESH MIAMI:
- South Florida’s AMPERA lands major shipping deal as it opens new global headquarters
- Ubicquia’s $106M raise signals a new phase for smart infrastructure
- Easymetering’s new smart panel is ready to outthink your home’s energy use
- Surfing the future of energy with NEPTUNYA
- How Ouros Energy is paving the way for a greener future
- Exowatt pulls in $50M to keep the AI boom from blowing a fuse
- As the World Cup arrives, Plei wants to get more people off the couch and onto the pitch - June 11, 2026
- With Minu acquisition, Miami-based unicorn Betterfly doubles down on U.S. growth - June 10, 2026
- Hallandale-based Lexful raises $7M seed round to reinvent IT documentation for the AI era - June 10, 2026