Recent outages underscore Anthropic’s ongoing infrastructure strain, with multiple elevated-error events and multi-model disruptions in August 2026—including a three-hour incident on August 24 affecting Opus 5, Mythos 5, and related services—reflecting demand growth that continues to outpace available compute. Trader sentiment clusters around 8–13 downtime days because historical patterns show several brief but recurring incidents per month, even as 90-day uptime hovers near 99.5 percent for core components. Differentiating factors include aggressive new capacity deployments via AWS, AMD, Nscale, and SpaceX partnerships, which could reduce peak-load failures, versus the risk that rapid model scaling and usage spikes will sustain volatility into September. The tight spread between leading buckets captures this balance between near-term reliability challenges and longer-term stabilization efforts.