Artificial intelligence (AI), particularly deep learning (DL) and large language models (LLMs), has rapidly gained prominence across industries. Training and serving deep neural networks (DNNs)—the core of LLMs—requires substantial computational resources, often provided by costly specialized hardware such as Graphics Processing Units. Cloud computing enables flexible access to such hardware but raises cost concerns. This study reports the significance of AI-related costs for Finnish software companies, based on a thematic subset of the 2025 Finnish Software Industry Survey. Of the 411 respondents, 64 reported developing or fine-tuning AI models, with 61% considering AI costs a significant or very significant concern. Respondents estimated AI-related expenses to rise from under 10% of cloud or hardware costs in 2025 to 10–25% in 2026 and 25–50% by 2028. These findings indicate that AI is expected to become a major driver of infrastructure costs, highlighting the importance of cost-efficient adoption strategies.