The article discusses a significant shift in the artificial intelligence sector, where the emphasis is moving from merely developing larger models to creating systems that optimize the use of various models based on specific tasks and cost considerations.
Perplexity CEO Aravind Srinivas highlighted that the product is no longer just the model itself but the orchestration system that integrates it with necessary tools and data. This change is driven by companies seeking cost-effective solutions as they tighten AI spending. Notably, open-weight models, which are more affordable and can be customized by companies, are gaining traction.
Benchmark general partner Peter Fenton predicts that over 90% of AI tokens will come from these open models within the next 18 to 24 months, potentially pressuring the profit margins of leading AI firms like OpenAI and Anthropic.
The article also notes that many Fortune 500 companies are adopting open models, which raises concerns about U.S. competitiveness as many of these models originate from Chinese labs.
The shift towards open models could lead to a hybrid AI system where routine tasks are processed locally, while more complex tasks are handled in the cloud, potentially altering the current demand for large data centers. Investors should consider how this evolving landscape may affect the pricing strategies and market positions of established AI companies