IBM (IBM) Bank of America Global A.I. Conference 2024 summary

AI strategy and product portfolio

• Focus on open, flexible AI solutions for enterprises, integrating RHEL AI for model deployment and OpenShift AI for scalable clustering of models.

• Expansion into business process integration through the watsonx portfolio, enabling enterprise data utilization and model orchestration.

• IBM Consulting accelerates adoption by providing expertise and implementation support for clients lacking in-house skills.

• Open source is central, with Granite models and InstructLab released under Apache License to foster innovation and commercial flexibility.

• Monetization spans GPU-attached RHEL AI, server-attached OpenShift AI, and business outcome-driven watsonx software, supported by OEM partnerships with Dell and Lenovo.

Market trends and enterprise adoption

• Enterprises face barriers in GenAI adoption: high costs, complexity of training on proprietary data, and limited flexibility for non-cloud use cases.

• Small, optimized models and hybrid infrastructure are emphasized to overcome these challenges and improve ROI.

• 2025 is anticipated as a pivotal year for enterprise AI production deployments, with infrastructure and technology readiness in place.

• Early adopters include biotech and industries with established data structuring, while most enterprises focus on automating business processes.

• Common use cases are support optimization, HR process automation, and operational improvements in procurement and discounting.

ROI, infrastructure, and competitive positioning

• ROI depends on reducing unit costs through small models and efficient infrastructure, with measurable gains in support and automation.

• Strategic acquisition of Neural Magic enhances model efficiency, enabling CPU-based inference and model sparsification.

• IBM's competitive edge lies in bridging foundation models with fit-for-purpose solutions, leveraging watsonx for business integration.

• Hybrid and multi-cloud architectures are increasingly prevalent, driven by data locality, sovereignty, and edge deployment needs.

• AI workloads are expected to shift toward inference over training, with IBM positioned to support both through flexible software and infrastructure.