Etched Raises $700M at $21B Valuation as Inference-Chip Demand Explodes

Skynet Chance (+0.02%): Cheaper, faster inference lowers the cost of running large numbers of autonomous model instances, modestly expanding the surface area for uncontrolled or unmonitored agentic deployment. The news contains no alignment or control mechanism to offset that.

Skynet Date (+0 days): A rapid capital influx into specialized inference silicon accelerates the buildout of deployment capacity, pulling forward the point at which powerful models operate at massive scale. The effect is on infrastructure pace, not capability discontinuity, so it is modest.

AGI Progress (+0.02%): Purpose-built prefill chips and cluster-scale shared memory attack the inference bottleneck that limits long-context reasoning and test-time compute, a currently important lever for capability gains. It is an engineering advance rather than a conceptual step toward general intelligence.

AGI Date (+0 days): Roughly $700 million more capital and a credible Nvidia alternative for inference increase the compute available per dollar, which shortens iteration cycles for reasoning-heavy models. Hardware still faces multi-year fab and deployment lead times, capping the acceleration.