We are the Frontier
U.S.

Open Weight AI Models Gain Favor in Enterprise Budgets

IBM pivot highlights a shift toward cost-efficient AI deployment as performance gaps between top models narrow significantly.

Updated on Oct. 22, 2026 in Artificial Intelligence

Open Weight AI Models Gain Favor in Enterprise Budgets

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 gaining enterprise traction by delivering 90% of frontier model performance at a fraction of the cost. The company sees this cost efficiency as a primary driver for shifting enterprise AI deployment strategies.

Why it matters

Enterprises are prioritizing cost-effective AI solutions over maximum performance, a trend that IBM is banking on to boost its infrastructure and software divisions. This shift suggests that businesses are moving away from top-tier proprietary models in favor of models that offer minimal performance trade-offs at significantly lower token costs.

Open weight models now operate at 10% to 25% of the token cost of frontier-grade models while maintaining 90% of their effectiveness. IBM is leveraging this efficiency gap to promote on-prem AI deployments via its Spyre mainframe accelerator hardware.

The players

IBM

A legacy enterprise technology conglomerate providing cloud infrastructure, software platforms, and mainframe hardware.

Arvind Krishna

The current CEO of IBM responsible for steering the firm's pivot toward hybrid cloud and AI-focused infrastructure.

Red Hat

An IBM-owned software subsidiary specializing in open-source enterprise solutions and Linux-based virtualization.

The details

IBM is positioning its software ecosystem to simplify the deployment of open weight models, which use shared weight parameters that developers can fine-tune for specific on-premises needs. By integrating these models with Spyre accelerator cards—specialized processors designed for AI-heavy workloads—the company aims to offset costs while maintaining local data sovereignty. This strategy contrasts with standard cloud-based frontier models that rely on continuous, expensive API-based token processing.

Timeline

  1. 2013: IBM acquired the Classic Cloud service.

  2. 2024: Red Hat virtualization products launched.

  3. Late 2025: Performance gap between models began shrinking.

  4. Q3 2026: IBM fiscal quarter concluded.

  5. October 22, 2026: IBM cloud maintenance initiated following critical PHP bugs.

The Tech Race

IBM's focus on open weight models marks a shift in the enterprise race to move AI workloads from expensive proprietary APIs to on-premises hardware. This strategy follows a clear trajectory from late 2025, when performance parity began challenging the dominance of high-cost frontier models.

Enterprise IT teams should anticipate easier integration paths for on-premises AI as IBM releases software to streamline deployment on Spyre hardware. Users of IBM cloud services may face temporary disruptions, with emergency maintenance protocols ongoing as of late October.

The takeaway

Enterprises are rapidly re-evaluating their AI spending, with open weight models currently capturing market share due to their clear cost-to-performance advantages. Watch the 2027 tech budget allocations to see if AI truly accounts for the projected one-fifth of enterprise spending.

Further reading

For more on the industry's shift toward accessible compute, see our Artificial Intelligence section.

Source note: This article includes information reported by TheRegister.

Live Poll

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