Date: September 4, 2026
Main source: The New York Times, republished by The Indian Express
Link: https://indianexpress.com/article/technology/artificial-intelligence/corporate-america-is-getting-hooked-on-open-source-ai-10865285/
Summary
Large US companies are increasing their use of open and open-weight models to cut costs and gain control over their AI systems. AT&T said these models already account for about 40% of its AI use and could reach 60%, with savings of up to 80% compared with certain closed alternatives. Airbnb and Deloitte also appear among the companies adopting this strategy. Proprietary models still have an edge on complex tasks, but open ones are gaining ground on specialized and repetitive workloads.
Why it matters
The enterprise market is moving toward multi-model architectures. If open models cover high-volume tasks at much lower costs, closed providers will have to justify their premium through superior capabilities, reliability or integrated products.
Practical implications
- Classify each workflow by difficulty and use premium models only when they produce a measurable improvement.
- Test open models for classification, extraction, summarization and repetitive internal processes.
- Measure cost per completed task, not cost per token in isolation.