A credible report suggests that the next generation of frontier models, likely GPT-6 or the next Anthropic model, will enable persistent, natively multi-agent systems capable of running for weeks, moving beyond current frontier lab capabilities to everyday users. This capability unlock is expected to drive significant revenue growth for AI labs, allowing for increased investment in compute and training for even better models.
The shift to persistent, multi-agent systems will create a new class of AI applications, driving substantial demand for advanced compute and training resources.