1. This Week in One Minute (Why This Matters)
The AI story is evolving.
For the last two years, investors focused primarily on models, software, and applications.
Increasingly, the market is shifting its attention toward the physical infrastructure required to
support artificial intelligence at scale. Energy, power delivery, semiconductors, industrial gases,
and data-center infrastructure are becoming the next battleground.
The companies best positioned today are not necessarily those building AI models. They are the
businesses providing the electricity, memory, power conversion, and infrastructure that make
those models possible.
At the same time, markets are becoming more selective. Investors are no longer rewarding
spending alone. They want evidence that capital-intensive investments will generate durable
returns.
2. Market Conditions We’re Monitoring
What we’re watching
- AI-driven electricity demand and grid capacity expansion
- Hydrogen infrastructure tied to energy security and industrial demand
- High-bandwidth memory (HBM) supply and pricing trends
- Capital expenditure discipline among hyperscalers
- Evidence that AI infrastructure investments are translating into economic returns
What would change our view
- Material easing of power or memory bottlenecks
- Significant deterioration in hyperscaler return profiles
- AI infrastructure spending accelerating without corresponding revenue generation
- Weakening demand signals across industrial and semiconductor infrastructure
3. Portfolio Lens (What This Means for Real Portfolios)
Markets are beginning to differentiate between AI excitement and AI economics.
Owning the infrastructure behind a technological transition often proves more durable than
attempting to predict which application ultimately wins. Whether through industrial gases,
memory, power conversion, or healthcare innovation, we continue to focus on businesses
positioned at critical points in their ecosystems.
Just as important, we remain focused on capital discipline. Attractive growth becomes far less
attractive when funded inefficiently.
4. High-Conviction Focus
A) AI Is Becoming an Infrastructure & Energy Story
The question we’re answering:
What happens when AI demand begins colliding with real-world infrastructure constraints?
What the evidence suggests:
Recent announcements from Google, Blackstone, Microsoft, Oracle, Bloom Energy, and Plug
Power all point toward the same conclusion: power availability is increasingly becoming a
limiting factor for AI deployment.
Google’s investment in virtual power plant capacity and broader AI infrastructure initiatives
represent some of the clearest evidence yet that hyperscalers are beginning to address energy
constraints directly.
Micron further reinforced the theme by indicating that 2026 HBM supply is effectively sold out
under long-term agreements, suggesting that memory demand is becoming increasingly
structural.
What we’re doing conceptually:
Maintaining focus on businesses exposed to electricity, power conversion, semiconductor
infrastructure, industrial gases, and other bottlenecks that support AI deployment.
B) Infrastructure Assets Are Replacing Narratives
The question we’re answering:
Which businesses benefit as energy security and infrastructure become strategic priorities?
What the evidence suggests:
Hydrogen is increasingly being viewed through a national security and energy resilience lens
rather than solely an environmental one.
Air Products remains a compelling example. Progress at NEOM, Louisiana Blue, and broader
industrial hydrogen adoption continues to support the long-term thesis. Equally important,
long-duration take-or-pay contracts provide stability through economic cycles.
At the same time, capital intensity remains a consideration. Large projects require patience,
disciplined execution, and thoughtful position sizing.
What we’re doing conceptually:
Favoring infrastructure businesses where long-duration demand is supported by contractual
economics and strategic relevance.
C) Capital Discipline Is Becoming the New Differentiator
The question we’re answering:
Can companies convert infrastructure spending into durable returns?
What the evidence suggests:
Microsoft continues strengthening its technology position through AI accelerators, proprietary
models, quantum advancements, and ecosystem partnerships.
Google continues investing aggressively in infrastructure and TPU development.
Vicor remains a niche beneficiary of AI power delivery demand, supported by backlog growth
and strong balance-sheet characteristics.
At the same time, investors are becoming increasingly sensitive to capital spending. Large
investments alone are no longer enough. Markets are demanding proof that these investments
generate attractive cash flows and durable economics.
What we’re doing conceptually:
Favoring businesses that can translate infrastructure spending into operating leverage, free
cash flow, and widening competitive advantages.
5. Weeding the Garden (Discipline in Action)
Continued monitoring of AI infrastructure beneficiaries where fundamentals remain
ahead of expectations.
- Maintaining discipline around valuation in areas where enthusiasm may be running
ahead of evidence. - Watching for opportunities created by volatility in capital-intensive infrastructure
businesses.
Patience remains an active decision.
6. Closing Perspective
The market narrative is evolving from “Who has the best AI model?” to “Who owns the
infrastructure required to power AI?”
That shift places greater importance on energy, industrial infrastructure, semiconductors,
power conversion, and disciplined capital allocation.
As always, our focus remains on identifying businesses where reality is beginning to exceed
expectations—not simply where excitement is greatest.