IBM CEO says open weight models are now irresistibly good and cheap on-prem

IBM’s CEO Arvind Krishna highlights a strategic shift in enterprise AI, noting that the performance gap between closed and open-weight models has narrowed significantly. While earlier assessments favored proprietary models, open-source options now deliver comparable effectiveness at a fraction of the cost. This drastic reduction in expense, combined with high efficiency, makes open-weight models a compelling choice for on-premises deployments, allowing businesses to handle the vast majority of daily tasks without incurring the premium prices associated with frontier models. The transition to on-prem AI infrastructure brings technical challenges, particularly regarding the specialized engineering skills required to manage GPU clustering and network optimization. To address these barriers, IBM is developing software solutions to streamline deployment processes. This strategic focus on accessible, cost-effective open models drives increased demand for local computing resources, influencing IBM’s hardware roadmap and reinforcing the importance of accessible data and model weights for enterprise adoption. This trend underscores the relevance of open data and open weights in the broader AI ecosystem, as enterprises seek to balance cost with capability. By favoring open alternatives, companies can maintain greater control over their data and infrastructure while reducing dependency on expensive cloud-based providers. The article illustrates how the democratization of high-quality AI models through open weights enables more sustainable and scalable enterprise implementations, shifting the focus from sheer model superiority to practical utility and economic viability.

Source: theregister.com
Published on 2026-10-23

IBM CEO says open weight models are now irresistibly good and cheap on-prem

Related news