Coverage Map

Browse research by topic.

Every InAI Capital research note is tagged across companies, technologies, and structural themes. Use this map to navigate the body of work — or to find the deepest piece on a question you're already chasing.

Companies

DeepSeek — V4 architecture deep dive
DeepSeek — V3 economics
DeepSeek — R2 multimodal
DeepSeek — V3.2 sparse attention
DeepSeek — mHC architecture
Alibaba Qwen — ecosystem
Alibaba Qwen — leadership
Baidu — ERNIE & Apollo
Moonshot — Kimi K2
ByteDance — DLCM

Hardware & Silicon

H100/H200/H800/B200 performance modeling
Nvidia → Huawei Ascend migration
H100 inference unit economics
INT4-native training (K2)
Ascend 910B at frontier scale
GPU rental & API pricing

Model Architecture

MLA dimension explosion & CSA/HCA
MoE & MLA at scale
Dynamic sparse attention
Manifold-constrained connections
Dynamic layer compression
INT4-native quantization

Robotics & Embodied AI

VLA East-West paradigm split
Humanoid BOM economics
Apollo Go robotaxi unit economics

Infrastructure & Agents

MCP & agent infrastructure
Inference profitability crisis
API pricing wars
Open-source enterprise stack

Policy & Geopolitics

Distillation & US-China transfer
Export controls & GPU supply
Hardware sovereignty
Open-source governance risk

Energy & Space Frontier

New coverage

The fourth pillar of the InAI Capital coverage map. Frontier intelligence, frontier energy, frontier space — the three are converging into one investable thesis. Research in this category is being scoped now; first notes ship in the coming months.

Compute power demand & grid economics Small modular reactors (SMR) Fusion investment landscape Launch-cost curves & reusability Satellite constellations & EO data In-space manufacturing & propulsion