“Intelligence is abundant. Returns are scarce. Fund the returns.”
Three signals define today's regime. Geopolitical friction is easing: Washington and Beijing agreed to reduce tariffs on $30 billion of non-sensitive trade in each direction and establish new channels for AI-risk dialogue. Yet strategic competition over advanced semiconductors, compute and technology sovereignty remains intact.
Capital has not repriced accordingly. The US 10-year Treasury closed above 5.15%, leaving long-duration AI infrastructure exposed to one of the highest hurdle rates of the cycle. Meanwhile, Middle East energy insecurity continues to embed a premium in power, logistics and inflation expectations.
The AI Supercycle remains intact. But the next phase will reward economic conversion, not technological ambition.
Global Rankings
No upgrades. No downgrades.
Five Signals
Where the regime is turning
-
The bilateral agreement reduces immediate trade friction and introduces mechanisms for managing AI incidents. But it does not alter the structural contest over semiconductors, compute and AI sovereignty. signal: Stability improves; productive capacity does not change.
-
A US 10-year yield above 5.15% materially raises the hurdle for long-duration AI investment. Capital remains available, but the investment test has hardened: CAPEX → UTILISATION → PRODUCTIVITY → CASH FLOW → ROIC > WACC. signal: The issue is no longer whether AI can attract capital — it is which AI deserves it.
-
The cost of intelligence increasingly extends beyond compute: COMPUTE + POWER + GRID + COOLING + FINANCING + RESILIENCE. The physical economy has re-entered the technology equation. signal: Power and resilience are now inputs to the AI cost curve.
-
The premium is shifting toward capacity that is POWERED → PERMITTED → CONTRACTED → UTILISED. Announced gigawatts without secured electricity, customers or financing deserve a widening discount. signal: Bankable capacity commands the premium; announcements do not.
-
Agentic AI is moving deeper into workflows, but adoption is no longer sufficient evidence of transformation. The next metrics are revenue per agent, productivity per employee, utilisation per accelerator and free cash flow per dollar of AI capital. signal: Deployment is table stakes; economics is the test.
The Hidden Pattern
Four regimes of the cycle
Board & Investor Agenda
From exposure to economics
Move to AI economics
Move from AI exposure to AI economics. Prioritise utilisation, pricing power, margins and free cash flow.
Maintain duration discipline
Higher sovereign yields materially reduce the value of distant AI cash flows.
Own scarcity that monetises
Memory, power, optics, networking, transformers, cooling and grids remain attractive where bottlenecks translate into pricing power.
Separate growth from value
AI revenue growth destroys value when incremental ROIC < WACC.
Board: capital-allocation agenda
Move AI decisively from the technology agenda to capital allocation. Require material investments to demonstrate CAPACITY → UTILISATION → PRODUCTIVITY → CASH FLOW → ROIC > WACC.
Board: measure and govern
Measure economic output per dollar of AI capital — not pilots, tokens or deployments. Treat power, cybersecurity, geopolitical exposure, financing and agent governance as board-level dependencies.
GVI 108.0, no ranking changes
- The AI Supercycle remains intact, but the market is moving from capacity to conversion.
- The next re-rating will reward companies that turn intelligence into productivity, cash flow and ROIC above WACC.
- Prioritise utilisation, pricing power, margins and free cash flow over headline exposure.
- Maintain duration discipline as higher yields discount distant AI cash flows.