Jordan Nanos

Overview

Jordan Nanos is presented in the source as an analyst at SemiAnalysis who studies AI compute, chips, data centers, and the economics of frontier-model development.

Current Profile

The bounded source positions Nanos as a market-oriented interpreter of AI safety and infrastructure demand. His central argument is that slowing model releases does not necessarily slow computing investment because evaluation, auditing, monitoring, reasoning, multi-agent execution, adoption of existing tools, and continuing research all consume capacity.

Key Characteristics

• Connects AI safety practice to concrete compute and infrastructure requirements.
• Separates pretraining, reinforcement learning, inference, adoption, and per-task reasoning as distinct demand drivers.
• Treats political opposition and regulation as more material downside risks than short-lived chip-market volatility.
• Expects efficiency improvements to be outweighed by expanded use, while acknowledging uneven bottlenecks and project timing.

Evidence

• Safety and controls: What could slowing the AI frontier mean for the economy? records Nanos linking evaluations, audits, controlled rollouts, network restrictions, training permissions, and monitoring to additional work and compute.
• Demand decomposition: What could slowing the AI frontier mean for the economy? distinguishes current-model adoption, reasoning, multi-agent workloads, reinforcement learning, and frontier research.
• Risk judgment: What could slowing the AI frontier mean for the economy? identifies regulation and public backlash as plausible threats and notes that some data-center projects may slip.

Qualifications

The profile rests on one short interview and does not independently establish Nanos’s forecasts, compute-share estimates, or claims about community benefits. The episode provides no underlying model, dataset, project list, or competing analyst view.

What Changed

• Established the initial source-bounded profile.

Relationships

• SemiAnalysis - organization through which the source identifies his research role.
• AI Safety Compute Demand - central economic mechanism he advances in the interview.
• Pacing the Frontier - policy debate whose infrastructure consequences he evaluates.
• AI Inference Cost Structure - broader economics of reasoning and agent workloads that his argument extends.