The Economics of the Bottleneck

“Velocity identifies the leader. The verdict identifies what can still break the thesis.”

The AI profit pool is migrating away from intelligence itself. As models proliferate, scarcity is moving along a chain — Memory → Packaging → Networking → Power → Chip Design → Security — while higher rates increase the cost of scaling every layer.

The emerging divide is Scarcity + Economic Capture versus Capital Intensity + Weak Conversion. Technology leadership remains essential, but the next phase will increasingly reward companies that can turn scarce positions into durable cash economics before the cost structure required to scale them absorbs the advantage.

The fundamental question: who can turn an indispensable position into durable economic value before capital intensity erodes the return?

1. 01 Broadcom | +0.6

   Evidence: proposed financing of up to $42 billion for Anthropic alongside expanding custom-silicon exposure. GVI impact: Capital Formation ↑ · Infrastructure Position ↑ · Ecosystem Strength ↑, partly offset by concentration risk. signal: Broadcom's role expands beyond semiconductor supply into infrastructure and customer economics.

2. 02 Micron | +0.3

   Evidence: $32 billion of long-term customer commitments amid constrained memory supply. GVI impact: Infrastructure Position ↑ · Capital Formation ↑ · Resilience ↑ · Execution Quality ↑. signal: Memory scarcity is becoming contracted and economically visible.

3. 03 Synopsys | +0.2

   Evidence: deeper OpenAI integration and an AWS licensing agreement above $1 billion. GVI impact: Innovation Velocity ↑ · Productivity Acceleration ↑ · Ecosystem Strength ↑. signal: Chip-design complexity is becoming a second-order AI profit pool.

4. 04 Anthropic | 97.5 | −0.5 | #11

   Evidence: at least $518 billion of long-duration infrastructure commitments. GVI impact: Innovation Velocity remains exceptional, while Resilience declines and the burden on future Economic Capture rises. Anthropic remains a Global Velocity Leader under absolute GVI thresholds but falls outside the relative Top 10. signal: Top-10 status is relative; GVI regime classification is absolute.

1. 01 The Economic Center of AI Is Moving Beyond Models

   Model capability is proliferating faster than the harder-to-reproduce layers surrounding it. TSMC captures advanced manufacturing scarcity. Micron captures AI-memory scarcity. Schneider Electric captures power architecture. Synopsys monetizes design complexity. Palo Alto Networks benefits from the governance burden created by autonomous agents. signal: AI value is moving toward the infrastructure and control layers required to industrialize intelligence.

2. 02 Resilience Is Becoming Strategic Power

   AI is becoming more capital intensive precisely as long-duration capital remains expensive. Companies with strong cash generation can invest through volatility, absorb delays and maintain strategic flexibility. Companies carrying large fixed commitments face a harder conversion test. signal: Higher-for-longer changes competitive structure — not merely valuation.

Five Takeaways

The decisive divide

• The economic center of AI is moving beyond models.
• Resilience is becoming strategic power.
• Funding provenance is becoming an investment variable.
• Leaders fund expansion; laggards disproportionately fund repair.
• The decisive divide is between strategic advantage already converting into cash and strategic advantage that still requires financial proof.