AI research and deployment, particularly in the field of generative AI, come with substantial computational costs. Many GenAI startups that train their own models are likely familiar with this challenge — you allocate at least half of your budget to GPU compute. What is the structure of the Total Cost of Ownership (TCO) of AI Infrastructure. How to reduce expenses? What are the hidden costs are standing behind the "unit price per GPU". This session will delve into the key components driving total costs in AI research and deployment and explore strategies to make your total cost of ownership (TCO) more predictable.