Gorilla Logic puts Latin America at the center of the debate as an AI innovation hub

Artificial intelligence (AI) is undergoing a shift in how companies perceive its costs, moving from a focus on training expenses to the expenses incurred during actual usage. This change is particularly prominent in Latin America, where software engineering company Gorilla Logic has been actively analyzing and discussing the implications.

According to Gorilla Logic's analysis, inference spending - the cost of using already-trained models in real operations - is projected to reach US$23.3 billion this year, surpassing training spending for the first time. The volatility of AI token costs, which can vary significantly depending on the task, has turned it into a new, unpredictable line item within the income statement.

This shift has caught many organizations off guard, with some exhausting their AI budgets within months. Gorilla Logic's engineers, spread across San José, Costa Rica, and Medellín, Colombia, have built a nearshore development infrastructure over a decade, taking advantage of the region's talent pool and technology ecosystems. The company now argues that controlling AI spending requires new approaches, such as using orchestration systems and gateways to assign tasks to cheaper models, and hybrid architectures that combine smaller, specialized models with more powerful ones.

While AI can potentially increase an engineer's productivity, the company emphasizes the importance of measuring the actual value generated by the technology, rather than just focusing on cost reduction metrics. The fundamental challenge for CFOs, CIOs, and companies in general is to move beyond simply managing AI costs and instead focus on quantifying the value that AI delivers to the business.