IBM CEO Touts Cost Benefits of Open Weight AI Models

Arvind Krishna says open weight models offer significant savings while maintaining high performance for enterprises.

Updated on Oct. 22, 2026 in Artificial Intelligence

IBM CEO Touts Cost Benefits of Open Weight AI Models

Live Poll

Would you prioritize using more affordable AI models if they performed similarly to expensive premium alternatives?

IBM CEO Arvind Krishna reports that open weight AI models are increasingly viable for enterprises due to their superior cost efficiency compared to frontier models. These models reportedly achieve 90 percent of the effectiveness of top-tier versions while costing just 10 to 25 percent of the token price.

Why it matters

Enterprises are increasingly prioritizing cost-effective AI solutions when the performance gap between models is minimal. IBM plans to leverage this trend by producing software to ease AI deployments and meeting expected on-prem demand with its Spyre accelerator cards.

Open weight models offer 90 percent effectiveness relative to frontier models at only 10 to 25 percent of the token cost. Meanwhile, IBM's infrastructure revenue grew five percent to $3.7 billion, while software revenue grew nine percent to $7.8 billion.

The players

IBM

This is a multinational technology corporation that provides hardware, software, and cloud services to global enterprises.

Arvind Krishna

He is the current Chief Executive Officer of IBM who oversees the company's strategic focus on AI and hybrid cloud computing.

Red Hat

This is an IBM subsidiary that focuses on open-source software solutions, including Linux and virtualization technologies for businesses.

The details

IBM is responding to the shift toward efficient AI by planning new deployment software and targeting on-prem demand for its Spyre mainframe accelerator cards. This strategic pivot comes as the company navigates a 20 percent decline in mainframe revenue, even as Red Hat segments show strong growth.

2013: IBM acquired the Classic Cloud service.
2024: Red Hat virtualization products launched.
Late 2025: Performance gap between AI models began shrinking.
Q3 2026: IBM fiscal quarter concluded.
October 22, 2026: Article published and IBM cloud maintenance ongoing.

The Tech Race

IBM's emphasis on open weight AI models positions the firm to capture value in the competitive on-prem infrastructure market. By integrating its Spyre mainframe accelerator cards, the company is attempting to outpace rivals by optimizing hardware for models that prioritize cost-efficiency over frontier-level scale.

Corporate users of IBM Cloud may experience service interruptions due to ongoing emergency maintenance lasting at least 16 hours. Customers should prepare for potential resource provisioning delays as the company addresses critical bugs in its software environment.

The takeaway

The pivot toward cost-efficient AI models suggests that enterprise users are increasingly focused on balancing performance with operational budgets. Companies can benefit by evaluating whether open weight models meet their specific needs as they plan for future tech expenditures.