Another reminder just hit the neocloud/AI trade: the biggest bottleneck is still compute supply. Moonshot AI's Kimi K3 launch saw demand spike so fast that new subscriptions had to be paused within days due to GPU capacity limits.

This isn't a direct update on any specific customer, but it reinforces the broader thesis. Frontier AI models can emerge anywhere, at any time, and immediately start consuming massive amounts of compute.

The AI race has shifted. It's not just about better models anymore; it's about who has the GPUs, data centers, and power to scale. Inference demand keeps growing faster than infrastructure can be built. More AI agents, more coding, more automation—it all just means more GPUs are needed.

That's the fundamental setup that benefits compute providers. Watching the entire AI infrastructure stack, including $NEBIUS(NBIS)$ , $IREN Ltd(IREN)$ , $CoreWeave, Inc.(CRWV)$ , and WYFI.