We are the Frontier
U.S.

Industry Giants Shifted Focus Toward AI Inference

As training demand levels off, companies are targeting a future where 90% of AI workloads consist of inference.

Updated on Oct. 13, 2026 in Artificial Intelligence

Bold flat-color editorial illustration of stacked geometric blocks representing server infrastructure, conveying the shift toward AI inference deployment.
IBM and Marvell Technology are reorienting infrastructure strategies toward AI inference as the industry pivots from model training to active AI deployment. AI Illustration. Upload story photo >

Live Poll

Do you plan to increase your investments in artificial intelligence technology this year?

IBM and Marvell Technology have outlined strategies focusing on AI commoditization and inference-heavy operations. These projections arrive as the industry shifts away from training-centric models, with current 60% training workloads expected to drop to 10%.

Why it matters

Companies are reorienting their infrastructure to capture long-term value from active AI use rather than model development. This transition is expected to prioritize return on investment as market attention turns to inference deployment.

Current AI industry workloads are split between 60% training and 40% inference. Future projections suggest a significant pivot toward a model where inference accounts for 90% of all processing demands.

The players

IBM

An enterprise technology company focused on hybrid cloud infrastructure and AI orchestration.

Marvell Technology

A semiconductor provider specializing in custom AI processors and optical networking.

Red Hat

An open-source software provider that serves as the foundation for IBM's AI infrastructure.

The details

IBM utilizes a four-part architecture consisting of agents, data, automation, and hybrid infrastructure to manage AI operations. The company leverages Red Hat—an open-source foundation for enterprise software—as the bedrock for its hybrid cloud deployment. By using an orchestration layer, IBM enables the integration of AI agents from multiple different vendors into a unified system.

Timeline

  1. The IBM Analyst Council meeting occurred in October 2026.

  2. Share prices for Nvidia, CoreWeave, Oracle, and Nebius Group declined last week.

  3. Marvell Technology announced a partnership with Nvidia's NVLink Fusion earlier this year.

  4. Marvell Technology expects to reach its revenue target in fiscal 2031.

The Tech Race

The transition toward an inference-heavy architecture represents a shift in competitive strategy away from massive model training. Major players are racing to provide the underlying infrastructure that commoditizes model use as primary revenue drivers.

For enterprise users, this shift implies that software stacks will rely more on integrated agent-based models rather than custom, internal training runs. Developers can expect infrastructure costs to eventually stabilize around deployment and active inference rather than the expensive initial model build.

The takeaway

The AI market is moving from the high-cost experimentation of model training toward the efficiency of operational deployment. Investors and engineers should watch for the actualized fiscal 2031 performance reports to see if these inference-driven revenue targets are met.

Further reading

For broader context on current hardware developments, see the latest updates in Artificial Intelligence.

Source note: This article includes information reported by SiliconANGLE.

Live Poll

Do you plan to increase your investments in artificial intelligence technology this year?