#infrastructure
4 posts
FERC Moves to Fast-Track AI Data Centers Onto the Grid: The Real Bottleneck Surfaces
A federal order pushing grid operators to connect AI data centers faster reveals the constraint behind the AI boom. It is not chips or models — it is power, and the wait to plug in.
AI Hardware in 2026: The Quiet Story Behind Cheaper Inference
The cheaper AI everyone is celebrating is partly a hardware story. NVIDIA Cosmos 3 and Intel Xeon 6+ are pushing the cost of running models down, and that changes more than benchmark scores.
Training a 100B-Parameter Model for $1.25 an Hour: AI's New Economics
Reports of a 100-billion-parameter model trained at roughly $1.25 per hour point to a real step-change in training cost. Here is what is genuinely new, what is hype, and what it means for builders.
Agentic AI Is Moving From Demos to Production, and Inference Is the New Bottleneck
Agentic systems are shifting from chat demos to real task completion, and the binding constraint is no longer model access but inference infrastructure. Here is what changes for teams.