Where the AI build-out binds next — and who owns the constraint
The question
What is the next company to grow 3x within 24 months on the combination of surging demand, a genuine supply bottleneck, monopoly-grade share, a growing industry, and capital inflows — identified before the move, from evidence available at the time?
How it works
The AI cycle is modelled as moving in waves — compute, then memory, then packaging, then optical, then power, then cooling, then the next constraint. Each candidate is scored against five pillars:
- Demand — is end-demand for the constraint actually surging?
- Bottleneck — is supply genuinely unable to respond?
- Monopoly — does one supplier hold a share the market cannot route around?
- Industry growth — is the industry itself expanding?
- Capital inflows — is money moving toward the constraint?
A candidate becomes a call only by passing the 3x Gate: ≥85/100 overall, with no pillar under 15. Evidence is primary-source first; figures that conflict across sources are held as UNVERIFIED and kept out of rankings. Every call is pre-registered and carries its own stated falsifier.
Cadence
Candidates are tracked and pillar-scored on daily runs. Calls, when the gate is passed, are scored at 6, 12, and 24 months. The module's first Update Study follows the first month of daily runs (late Nov 2026), or earlier if a candidate passes the gate.
Current status — Oct 26, 2026
The founding study is complete. 13 candidates are tracked; 2 further names were examined and explicitly classified as not bottlenecks. The earliest-stage candidate is ABF film / T-glass: Ajinomoto holds >95% of ABF film, Nittobo ~90% of T-glass. On energy, the shortage is real but merchant nuclear is assessed late / post-peak, grid equipment mid-cycle, and regulated utilities early-mid with rate-capped returns. No candidate has passed the 3x Gate yet — so no calls have been made.
Studies
- SIP-2026-001 — The Bottleneck Playbook · Founding Study · Oct 26, 2026
Scorecard summary
0 gate passes · 0 live calls. Selectivity is the strategy — the record so far is a tracked candidate list, not a set of calls. See the full scorecard →