Source: The Register: Biting the hand that feeds
After years of betting on proprietary AI stacks, IBM's leadership is publicly admitting that open-weight models deliver comparable performance at a fraction of the cost—a significant concession that accelerates enterprise defection from closed platforms. The on-premise deployment model these open models enable removes a key lock-in mechanism: API dependency and cloud vendor pricing. Enterprises can now run sophisticated inference locally without negotiating with cloud providers. IBM's own admission matters less than what it signals: even the companies who built the original closed-model playbook no longer believe they can defend it on economic grounds.