Fast takeoff
Capability outruns deployment, power and regulation.
Compute ↑ · Grid ↑Investment doctrine
One thesis. Three scenarios. Eight independent ways to be right.
We do not predict an AGI date. We own scarce assets needed across plausible trajectories and adjust exposure when evidence changes.
From intelligence to electrons
From intelligence to electrons
Each layer has its own economics, cycle and risk. Together they reduce dependence on a single winner.
01
Models + cloud
38%Frontier models, distribution and enterprise demand
02
Compute
17%Accelerators, custom silicon and software ecosystems
03
Semicap + memory
18%Fabs, EUV, yield, packaging and HBM
04
EDA + IP
3%Critical software and IP for complex chip design
05
Networking
5%AI fabrics, optics and high-speed interconnect
06
Power + grid
14%Electricity, cooling, generation and construction
07
Data centers
1.5%Interconnection-rich locations and deliverable power
08
Reserve
3.5%Liquidity for drawdowns, costs and opportunities
Industrial scaling
Models improve steadily while value moves into inference and infrastructure.
Balanced stackPlateau / regulation
Progress or access slows; cash-flow quality and liquidity matter more.
Quality ↑ · Reserve ↑01ObserveFilings · IR · policy · supply chain
02FalsifyBear case · base rates · source quality
03ProposeWeight · band · reason · expiry
04AuthorizeHuman IC · broker controls · audit log