The constraint investors have been working through all summer is where AI inference actually runs: on a device or through a cloud-based GPU cluster.
Running an open-weight language model requires dedicated GPU compute. That hardware requirement is what commercial API arrangements sidestep, and it is now where Latham & Watkins is putting capital.
JPMorgan has terminated its lending relationship with Situational Awareness, the hedge fund founded by Leopold Aschenbrenner, after the fund absorbed billions of dollars in losses during an AI sell-off
The capability threshold that has converted AI safety concerns from academic footnote to industry headline is the autonomous agent.