AnthonieMeijer
@anthoniemeijer.bsky.social
Writing on AI & investing | VC investor | Ex-strategy consultant
https://anthoniemeijer.substack.com/
https://anthoniemeijer.substack.com/
Investment filter for AI infrastructure:
Don't ask "what chips do they use?"
Ask "what's their power situation?"
The grid is the new moat.
Don't ask "what chips do they use?"
Ask "what's their power situation?"
The grid is the new moat.
January 18, 2026 at 11:15 AM
Investment filter for AI infrastructure:
Don't ask "what chips do they use?"
Ask "what's their power situation?"
The grid is the new moat.
Don't ask "what chips do they use?"
Ask "what's their power situation?"
The grid is the new moat.
A single hyperscale AI site needs 30,000-40,000 tons of copper.
GE Vernova: +103% return in 2025. Backlog: $135B+
The chip trade was 2023.
The power trade is 2025.
The copper trade is AI exposure without model risk.
GE Vernova: +103% return in 2025. Backlog: $135B+
The chip trade was 2023.
The power trade is 2025.
The copper trade is AI exposure without model risk.
January 18, 2026 at 11:15 AM
A single hyperscale AI site needs 30,000-40,000 tons of copper.
GE Vernova: +103% return in 2025. Backlog: $135B+
The chip trade was 2023.
The power trade is 2025.
The copper trade is AI exposure without model risk.
GE Vernova: +103% return in 2025. Backlog: $135B+
The chip trade was 2023.
The power trade is 2025.
The copper trade is AI exposure without model risk.
Fundamentals didn't matter in early crypto.
They matter more now.
They'll matter a whole lot more in the future.
Which protocols pass your quality filter?
They matter more now.
They'll matter a whole lot more in the future.
Which protocols pass your quality filter?
January 16, 2026 at 5:55 PM
Fundamentals didn't matter in early crypto.
They matter more now.
They'll matter a whole lot more in the future.
Which protocols pass your quality filter?
They matter more now.
They'll matter a whole lot more in the future.
Which protocols pass your quality filter?
Protocols that actually pass:
- Aave: $157M revenue, $50M/yr buybacks, $25B loan book
- Hyperliquid: 97% of fees → automated buybacks ($645M YTD)
- GMX: 12.9% of supply retired, 30% fees to stakers (at ATL now)
Real cash flows. Not narratives.
- Aave: $157M revenue, $50M/yr buybacks, $25B loan book
- Hyperliquid: 97% of fees → automated buybacks ($645M YTD)
- GMX: 12.9% of supply retired, 30% fees to stakers (at ATL now)
Real cash flows. Not narratives.
January 16, 2026 at 5:55 PM
Protocols that actually pass:
- Aave: $157M revenue, $50M/yr buybacks, $25B loan book
- Hyperliquid: 97% of fees → automated buybacks ($645M YTD)
- GMX: 12.9% of supply retired, 30% fees to stakers (at ATL now)
Real cash flows. Not narratives.
- Aave: $157M revenue, $50M/yr buybacks, $25B loan book
- Hyperliquid: 97% of fees → automated buybacks ($645M YTD)
- GMX: 12.9% of supply retired, 30% fees to stakers (at ATL now)
Real cash flows. Not narratives.
The crypto quality filter:
✓ Network effects or switching costs
✓ Protocol revenue > token emissions
✓ Fee capture (buybacks, burns, yield)
✓ Aligned, active team
Sound familiar? It's just ROIC + moat + management, translated to on-chain.
✓ Network effects or switching costs
✓ Protocol revenue > token emissions
✓ Fee capture (buybacks, burns, yield)
✓ Aligned, active team
Sound familiar? It's just ROIC + moat + management, translated to on-chain.
January 16, 2026 at 5:55 PM
The crypto quality filter:
✓ Network effects or switching costs
✓ Protocol revenue > token emissions
✓ Fee capture (buybacks, burns, yield)
✓ Aligned, active team
Sound familiar? It's just ROIC + moat + management, translated to on-chain.
✓ Network effects or switching costs
✓ Protocol revenue > token emissions
✓ Fee capture (buybacks, burns, yield)
✓ Aligned, active team
Sound familiar? It's just ROIC + moat + management, translated to on-chain.
Some examples of AI-native SaaS with real ROI:
- Shopify: 90% dev adoption, 24K lines AI code accepted/day
- Duolingo: 92% faster course creation, 67% faster code reviews
- Synthesia: 98% self-serve during 690% volume spike
However, the question remains whether this affects the P&L as well
- Shopify: 90% dev adoption, 24K lines AI code accepted/day
- Duolingo: 92% faster course creation, 67% faster code reviews
- Synthesia: 98% self-serve during 690% volume spike
However, the question remains whether this affects the P&L as well
January 9, 2026 at 5:46 PM
Some examples of AI-native SaaS with real ROI:
- Shopify: 90% dev adoption, 24K lines AI code accepted/day
- Duolingo: 92% faster course creation, 67% faster code reviews
- Synthesia: 98% self-serve during 690% volume spike
However, the question remains whether this affects the P&L as well
- Shopify: 90% dev adoption, 24K lines AI code accepted/day
- Duolingo: 92% faster course creation, 67% faster code reviews
- Synthesia: 98% self-serve during 690% volume spike
However, the question remains whether this affects the P&L as well