IBM CEO says businesses are choosing open-weight models over expensive frontier AI
IBM CEO Arvind Krishna says enterprises increasingly prefer open-weight AI models, citing comparable performance at a fraction of the cost of frontier models.Summary
On IBM's Q3 2026 earnings call, CEO Arvind Krishna said that enterprises are increasingly choosing open-weight AI models over expensive frontier LLMs from companies such as Anthropic and OpenAI, because they offer roughly 90% of the performance at about a tenth of the cost. He noted that the performance gap between top frontier models and open alternatives has narrowed sharply, from a 70-versus-20 or 30 benchmark spread in 2024 to closer to 70-versus-50 in late 2025, and he expects it to reach roughly 100 versus 90. Krishna added that a typical large enterprise can run open-weight models with just two or three racks of infrastructure, at 10% to 25% of the token cost of frontier models, making them highly attractive for most tasks. He also highlighted AI sovereignty as a growing concern, since on-premises and private-cloud deployment of open-weight models gives enterprises permission and IP rights to control their data and logs rather than leaving traces with large public LLM providers.
(Source：Hotminute)