SpaceX Builds In-House Turbine Blade Foundry to Break AI's Power Bottleneck

Skynet Chance (+0.02%): Removing a hard physical constraint on data center power expands the total compute available for frontier training runs and concentrates a critical supply chain capability under a single actor, marginally increasing the odds of rapid, weakly-governed capability scaling. The effect on loss-of-control risk is indirect, working through capacity rather than through alignment or autonomy.

Skynet Date (+0 days): Cutting up to 18 months off turbine deployment timelines directly compresses the schedule on which large new compute clusters can be energized, pulling any capability-driven risk milestones somewhat earlier. Vertical integration by a well-capitalized AI builder also reduces the natural pacing effect that supply oligopolies impose.

AGI Progress (+0.01%): Energy availability has emerged as a binding constraint on scaling alongside chip supply, so unlocking turbine production capacity meaningfully raises the ceiling on deployable training and inference compute. It is an enabling infrastructure step rather than an algorithmic or theoretical advance toward general capability.

AGI Date (+0 days): An 18-month acceleration in bringing power online, against IEA projections of data center electricity use doubling by 2030 and GE Vernova being sold out through 2030, could pull forward the timeline for the largest training runs. Execution risk is real, since single-crystal casting at power-plant scale is difficult, and pollution litigation could slow deployment regardless of manufacturing throughput.