ISSUE 003 · The Aroko Letter
Orchestrating the Agentic Power Surge
What continuous, agentic AI workloads actually ask of the grid — as distinct from training bursts
Training is a burst; inference at scale never stops. This issue is the earliest statement of the load-profile shift that later issues keep returning to — planning for a permanent new customer, not an occasional spike.
Key takeaways
- The distinction between training (burst) and inference (continuous baseload) load reappears as a load-bearing assumption in nearly every later issue.
READ THE FULL ISSUE Published in full, free, on Beehiiv.
Forecasts this issue put on the watchlist
Open Grid Connections & Local ApprovalCPower/Bentaus production deployment results.
Target Q2–Q3 2026 — operator disclosures, CAISO/ISO-NE data
Open Grid Connections & Local ApprovalKraken–Uplight integration timeline for dispatching AI compute loads in live ISO markets.
Octopus Energy / Kraken quarterly updates
Open FinancingNemoClaw/OpenClaw enterprise adoption velocity among GTC 2026-named enterprises.
Enterprise earnings calls, software procurement data
Open Grid Connections & Local ApprovalFERC demand-response classification for AI compute.
Target Q3–Q4 2026 — FERC docket filings
Open Power EquipmentGE Vernova and Siemens Energy backlog.
Quarterly earnings calls
The Aroko Letter
Free. Biweekly. No sponsored research. Every forecast tracked in public with its kill condition attached.
GET THE LETTER