The AI Capital Rotation

“Follow the returns, not the spending.”

AI leadership is shifting from technological scale to return on capital. The strongest companies are not simply spending the most. They combine control of critical infrastructure, financial resilience and improving returns on invested capital.

Amazon's proposed chip-financing structure is today's clearest signal of that transition. If AWS can separate access to compute from ownership of rapidly depreciating GPUs, it may scale capacity while preserving capital for higher-return infrastructure. The benefit is real only if lease economics improve total returns rather than merely transform capex into fixed obligations.

At the same time, companies controlling non-substitutable parts of the AI stack — advanced manufacturing, memory, power, networking and design infrastructure — retain structural leverage. The investment conclusion: favor rising Structural Velocity where ROIC and free-cash-flow conversion are improving; raise the hurdle where spending, leverage or valuation are rising faster than returns.

1. 01 Amazon — GVI 99.7 (+0.1, ↑↑)

   Amazon is exploring a structure under which approximately $8 billion of Nvidia chips could be owned by external investors and leased for AWS use. It has also committed more than $1 billion over five years to communities surrounding its U.S. data-center footprint. The transmission lifts Capital Formation, Strategic Optionality and Resilience, offset by rising Financing Complexity and Lease/Counterparty Exposure.

   The signal is not lower AI investment — it is potentially more efficient financing of that investment. If Amazon can separate compute usage from compute ownership, AWS could preserve capital for data centers, networking, power infrastructure and proprietary silicon while continuing to deploy Nvidia capacity at scale. The test is specific: does the structure lower the effective cost of compute and improve incremental ROIC, or simply move capex into lease obligations? signal: Until returns are demonstrated, the signal supports strategic optionality — not superior returns. Score: 99.6 → 99.7.

2. 02 Broadcom — GVI 98.5 (+0.1, ↑↑)

   A major financing package is being assembled around AI-chip deployment across Broadcom-linked customers, with Broadcom potentially providing residual-value support on part of the senior financing. The move lifts Capital Formation, Infrastructure Position, Ecosystem Strength and Strategic Optionality, offset by rising Customer Concentration and Contingent Financial Exposure.

   Broadcom is evolving beyond semiconductor supply toward a broader role in the financing architecture surrounding custom AI infrastructure. That deepens ecosystem control while increasing exposure to customer economics. signal: Supplier power is becoming financial power. Score: 98.4 → 98.5.

3. 03 Anthropic — GVI 97.4 (−0.1, #11 overall)

   Anthropic remains a Global Velocity Leader because its score exceeds the 97.0 threshold, but it ranks #11 overall because the current #10, Alibaba, scores 98.1. Innovation Velocity and AI Adoption remain exceptional, but Resilience and Strategic Optionality are falling while Execution Risk rises alongside expanding infrastructure, financing, regulatory and geopolitical exposure.

   Anthropic's challenge is increasingly financial rather than technological. Can revenue scale faster than the fixed infrastructure commitments required to produce it? signal: Structurally high-velocity, but below the Top-10 cut line. Score: 97.5 → 97.4.

4. 04 TSMC — GVI 99.9 (0.0, ↑)

   TSMC continues to evaluate additional U.S. manufacturing options beyond its major Arizona investment program. No additional Texas commitment has yet been finalized. signal: Optionality rises; Capital Formation does not yet. No score change.

Five Market Takeaways

The return-on-capital lens

• AI is now a return-on-capital trade. Capex is no longer the signal — ROIC and free-cash-flow conversion are.
• Bottleneck control is becoming pricing power. Advanced manufacturing, memory, power, networking and chip design are the AI economy's toll roads.
• AI financing is becoming a competitive weapon. Separating asset ownership from asset usage can preserve flexibility — but only if financing costs stay below the value created.
• Higher-for-longer rewards financial endurance. Strong balance sheets let leaders keep investing while weaker competitors retrench.
• Cash conversion is the final test. Technology leadership must ultimately produce durable earnings, free cash flow and superior returns on capital.