IBM CEO: open-weight models are now good and cheap to run on-prem

IBM's chief executive has weighed in on the shifting balance of power in the AI market. In his view, open-weight models have matured to the point where they can no longer be dismissed: they are now both capable and inexpensive to run on a customer's own servers.

The central claim is the appeal of local deployment, where a model runs inside the company's own infrastructure (on-prem) rather than in a third party's cloud. This approach has long appealed to organizations that value control over their data, predictable costs, and independence from outside services. The IBM chief's argument is that the combination of improved open-model quality and falling operating costs makes this option close to irresistible for business.

The stance is notable coming from a company that itself sells AI solutions to enterprises and has deep roots in working with large organizations. Betting on open weights and local execution differs from the strategy of vendors who emphasize proprietary models available only through the cloud.

The piece drew considerable attention on Hacker News, gathering 129 points and 80 comments — a sign that the question of open versus closed models, and of cloud versus on-prem, remains a lively debate across the industry.

Source: Hacker News