Will an AI datacenter in the US be sabotaged before 2029?
βž•
Plus
8
αΉ€594
2029
51%
chance

This question resolves yes if there is credible reporting of an incident of sabotage, between September 2024 and December 2028 inclusive, of a datacenter used for AI training or inference in the US (or dedicated power infrastructure for such a datacenter) resulting in either $100M+ in physical damages to the datacenter (and/or its dedicated power infrastructure), or a 2+ month delay in the facility's effective use of at least 50,000 H100 (or better) GPUs.

In particular, for an incident to count as sabotage, the following must be widely believed:

  • The incident was the result of a premeditated, malicious action by some party, rather than an accident.

  • Damaging AI-relevant infrastructure was a primary motivation of the incident, such that the damage to AI-relevant infrastructure appeared to be targeted rather than incidental. For example, large-scale attacks on critical infrastructure wouldn't count if there is scant evidence that the responsible party was not primarily interested in damaging AI-relevant infrastructure. To determine whether this condition is met, I'll gauge the extent to which reporting characterizes AI-relevant infrastructure as the target of such an attack.

Get αΉ€1,000 play money
Sort by:

I can't wait for LLMs to become actually good at operationalizing these.