Practical intelligence on the developer-to-operator transition—operational readiness, staffing ramp, turnover risk, OT governance, and the evidence that protects your SLAs.
Where inference capacity actually gets built, what it costs, and why the 5–20 MW metro site is becoming the unit that matters.
CBRE's H1 put asking-rent growth for 3–10 MW deployments at 8.3%, the fastest of any size band. Inference is changing who gets the best pricing.
Read the piece →A B200-hour rose 20.5% in September while OpenAI halved GPT-6 Sol's list price. That gap needs 2.4x more tokens out of every GPU-hour — and the racks that deliver it draw 190–230 kW.
Read the piece →PJM and NERC propose large-load rules starting at 50 MW; Dominion's GS-5 starts at 25 MW. Below a cutoff those obligations don't apply — but the site still has to check out. The 12 questions we ask first.
Read the piece →Ready for Inference runs on LinkedIn. New editions post there first, and the comments are where the interesting disagreements happen.
Insights are useful. An engagement that turns them into artifacts, governance rhythms, and evidence trails is what protects your SLAs.