Strategic Intelligence Report

The State of AI

Orchestrating Intelligence in the Age of AI
March 2026 — Full Report Edition
22
Chapters
6
Parts
22
Interactive Dashboards
1,000x
Intelligence Yield
Why this report exists. The convergence of declining AI costs, exploding capabilities, and enterprise adoption creates a once-in-a-generation strategic inflection point. Intelligence per dollar has improved 1,000x since GPT-3 — following a predictable compound curve that shows no sign of slowing. This report provides the data, frameworks, and playbooks to act on it. Each chapter contains narrative analysis woven around interactive charts and data visualizations. Click any chapter to read the full analysis.
Executive Summary The State of AI in Five Minutes
The complete thesis distilled: intelligence per dollar has improved 1,000x, infrastructure will commoditize, and value migrates to the application layer. Key findings, strategic implications, and the action framework — all in one page.
Read the Executive Summary
Part I The Intelligence Layer
What AI can do and how fast it's evolving
Ch 1 The Intelligence Yield Hypothesis
The core thesis — intelligence per dollar has improved 1,000x since GPT-3, following a predictable compound curve driven by three vectors: architectural efficiency, price competition, and algorithmic amplifiers.
Interactive Dashboard
Ch 2 The Intelligence Cost Curve
How AI pricing is commoditizing faster than any technology in history. The "Densing Law" — capability density doubles every 3.5 months.
Pricing & Commoditization Dashboard
Ch 3 The Great Convergence (Gap Evolution)
The gap between frontier and open-source models has collapsed from 24+ months to under 6. Why this convergence reshapes competitive dynamics.
Gap Evolution 2022–2026
Ch 4 Model Tiers (S/A/B/C)
A comprehensive tier-by-tier comparison of frontier, near-frontier, and commodity models as of February 2026.
Model Tier Comparison
Part II The Hypothesis — Value Migrates to Applications
The central prediction: infrastructure commoditizes, applications capture value
Ch 5 AI — The Current Revolution Unfolding
91% infrastructure today. History says this inverts to 20/80 within a decade. NVIDIA is in the exact position of Edison, AT&T, and Intel.
Deep Dive
Ch 6 The Value Layer Shift
The universal pattern: infrastructure commoditizes, value migrates to the application layer. Four revolutions, one pattern, one prediction for AI.
Cross-Revolution Overview
Part III Historical Analogies
Proving the pattern with 140 years of evidence
Ch 7 Electricity — From Edison's Empire to the $28T Enabled Economy
Edison controlled the entire stack. 120 years later, utilities capture 1.8% of GDP. The original infrastructure-to-application value migration.
Deep Dive
Ch 8 Telecom — From AT&T's Empire to Netflix's Dominance
AT&T spent $133.5B trying to escape the dumb pipe fate. OTT applications are now worth 5.7x the infrastructure. The crossover happened in 2020.
Deep Dive
Ch 9 Silicon — From Intel's Dominance to Microsoft's 13x Lead
Equal partners in 2000. By 2026, software is worth ~13x silicon. The Wintel dissolution and the shift from chips to cloud.
Deep Dive
Ch 10 The Britannica Problem
Every new model spends $100M+ re-learning Wikipedia. The diminishing returns of scale and the historical parallel.
Re-Learning Waste Analysis
Part IV The Model Landscape
Understanding the tools at your disposal
Ch 11 Model Taxonomy
A framework for matching model capabilities to task requirements — from sub-$0.001 micro-tasks to $2+ frontier reasoning.
Interactive Taxonomy
Ch 12 The Small Model Thesis
Why 80% of enterprise AI workloads don't need frontier models. The economics of small, specialized models that deliver 95% of value at 1% of cost.
Enterprise Task Automation
Ch 13 Visual & World Models
The state of the art in multimodal AI — image, video, 3D, and world simulation. Where visual intelligence stands and where it's heading.
Multimodal AI Dashboard
Ch 14 Model Training Economics
From Chinchilla scaling laws to $500M+ frontier training runs. Why training costs are exploding while inference costs collapse.
Cost Analysis Dashboard
Ch 15 The Open Source Paradox
Why Meta gives away models that cost $100M+ to train. The strategic calculus of open vs. closed, and how this reshapes the entire industry.
Strategic Analysis
Part V The Infrastructure Arms Race
The physical foundations of the intelligence revolution
Ch 16 GPU Compute Demand
$1.7 trillion in compute infrastructure. How the AI buildout maps to GPU demand and the physical limits of scaling.
$1.7T Compute Stack
Ch 17 NVIDIA Infrastructure Analysis
NVIDIA's dominance and the physical limits of the AI buildout. Infrastructure projections 2025–2030.
NVIDIA Infrastructure 2025–2030
Ch 18 Inference Demand
296 trillion tokens per day and what it means for infrastructure, cost, and the shape of the AI economy.
296T Tokens/Day Analysis
Part VI The Enterprise Disruption
Which industries, which jobs, what timeline, and how to build
Ch 19 The $600B Disruption Map
Enterprise AI is not a technology upgrade — it's an industry restructuring. Mapping which sectors face disruption, the magnitude of impact, and the timeline.
Interactive Disruption Map
Ch 20 SaaS Disruption Timeline
Year-by-year projection of how AI transforms the $300B SaaS industry. From copilots to agents to autonomous workflows — which categories fall first.
Year-by-Year Transformation
Ch 21 Task Mapping
A granular analysis of enterprise job functions, the tasks within them, and how each maps to specific AI capabilities and cost profiles.
Interactive Task Mapping
Ch 22 The Intelligence Routing Revolution
Nobody runs the frontier model for every query. The real architecture is tiered routing — 70% tiny agents, 20% mid-tier specialists, 10% frontier reasoning.
Routing Architecture
Conclusion The Intelligence Advantage Belongs to Those Who Orchestrate
Twenty-two chapters. Six parts. One thesis: intelligence per dollar is compounding at 1,000x, infrastructure will commoditize, and value will migrate to the application layer — exactly as it did with electricity, telecom, and silicon.
Read the Conclusion

Appendices

A: Methodology & Validation — Intelligence Index definition, benchmark selection, data sources
B: Model Reference Cards — Frontier model specs, pricing, capabilities (Feb 2026)
C: Glossary — Intelligence Yield, Densing Law, Intelligence Routing, Value Layer Shift