The 80% Problem: Why Most Enterprise AI Doesn't Need a Frontier Model | Hyperstack Panel & Networking

The 80% Problem: Why Most Enterprise AI Does Not Need a Frontier Model The 80% in the title is not a Hyperstack statistic. Most enterprise AI programmes still treat the largest available model as the default for every task. This panel will argue that production work splits: open-weight models for the bulk of the load, the frontier for the narrow set of tasks that need it, with the weights, the data, and the deployment inside the customer's own perimeter. Discussion topics Where open weights win, and where they do not. What the hidden costs look like in MLOps, evaluation, and model lifecycle. What control buys an enterprise beyond a line in a compliance file. Followed by drinks, food and networking at Bellboy Bar - one of Berlin's most distinctive cocktail bars. The panel Rod Evans, EMEA VP - Supercomputing & AI Cloud Infrastructure, NVIDIA Gareth Mitchell-Jones, Head of AI, Daemon Darren Franklin, Chief Technologist for Data and AI, Computacenter Agenda 18:30 Doors open & welcome drinks 19:00 Hyperstack intro & panel session 19:30 Drinks, food, and networking until late Please join us if you're an AI leader, work in platform or infrastructure, or you're at GTC and just curious where it's all headed. About Hyperstack Hyperstack is an NVIDIA Cloud Partner. On-Demand GPU cloud and Secure Private Cloud for teams that need the model and the data to stay under their control.