ABAB Pedia

Max Kan

Max Kan is a SemiAnalysisSemiAnalysisSemiAnalysis is a semiconductor and AI research firm founded by Dylan Patel in 2020. It produces reporting, institutional data models, consulting and hardware testing, covering chip manufacturing, memory, datacenters, inference costs and AI economics.Open the full entry analyst and co-lead of its Tokenomics team. He studies AI token demand and inference economics, holds a computer-science degree from Williams College, previously worked in Strategy and Product at Jane Street and co-founded reinforcement-learning environment startup Habitat.

Contents9 sections
Key facts

Williams College

Kan studied at Williams College from 2020 to 2024 and earned a bachelor’s degree in computer science. His public professional profile also lists him as valedictorian.⁠[1][2]

internships and Jane Street

His public résumé lists a Google software-engineering internship in summer 2022 and a Jane Street Strategy and Product internship in summer 2023. He held a Strategy and Product role at Jane Street from September 2024 to May 2025.⁠[3]

co-founding Habitat

Kan co-founded Habitat, a startup developing reinforcement-learning environments. His public résumé dates his co-founder tenure from May 2025 to January 2026; that personal employment record does not establish the company’s subsequent status.⁠[1][3]

joining SemiAnalysis

Kan’s public résumé dates his start as a SemiAnalysis analyst to April 2026. The firm’s biography identifies him as co-lead of Tokenomics, which studies relationships among compute capacity, token supply and AI business revenue.⁠[3][1][4]

subscription-economics report

SemiAnalysis Weekly appearance

Kan appeared with Andrew Megalaa and Jordan Nanos in episode 36 of SemiAnalysis Weekly, covering subscription testing, caching, usage limits and token efficiency on the firm’s own channel.⁠[6]

Topics: public discussions of inference architecture

A WEKA event lists Kan as SemiAnalysis’s Tokenomics technical lead alongside WEKA chief AI officer Val Bercovici and Oracle architect Manu Mishra. Its topic is CPU, GPU and KV-cache behavior in agentic inference. The vendor-event description does not establish a universal architecture conclusion.⁠[7]

Public profile and accounts

Sources

  1. Max Kan — Analyst and Tokenomics co-lead (opens in a new window)
  2. Max Kan — education (opens in a new window)
  3. Max Kan | LinkedIn (opens in a new window)
  4. Tokenomics Model (opens in a new window)
  5. Anthropic Subscriptions Offer 5x+ More Value Than OpenAI (opens in a new window)
  6. SemiAnalysis Weekly, Episode 36 (opens in a new window)
  7. Does Agentic AI Break the Case for GPU-Only Inference? (opens in a new window)
  8. Max Kan | X (opens in a new window)