G – Growth Engine
I – Industry Position
D – Durability of Moat
D – Downside Protection & Risks
A – Alignment with Shareholder Value
You’ll get the straight-talk Buffett/Munger-style view, plus a new mental model at the end to think about AI-infra companies differently.
🟦 GIDDA on NBIS (Nebius Group)
G — Growth Engine
Score: Strong → Very Strong
What’s driving it:
- They sit in the picks-and-shovels of AI: compute, GPU clusters, specialized cloud infrastructure.
- Demand is surging + chronic global GPU shortage.
- They’ve been doing 300–350%+ YoY revenue growth in multiple quarters.
- Their partnership with NVIDIA is real and critical.
- They are positioning as a Europe-friendly, privacy-friendly alternative to US hyperscalers.
- High-capacity datacenter buildout pipeline underway (multi-country).
Skeptical check:
- Growth is somewhat supply-constrained, not demand-constrained.
- If supply loosens (more H200 / B200 availability), margins may compress.
- AI infra is capital-intensive: capex eats operating leverage.
Verdict:
Growth engine is firing hard; massive tailwinds. But very capex-dependent and cyclical.
I — Industry Position
Score: Moderate → Strong
Strengths:
- They operate where regulation + data sovereignty matters — EU, MEA, India-like markets.
- Positioned as an elastic, GPU-focused alternative to AWS/Azure/GCP.
- Strong developer adoption in niches (training workloads, EU data compliance).
Weaknesses:
- Competing against trillion-dollar giants with cheaper capital.
- Brand is not globally known.
- Relying heavily on NVIDIA — dependency risk.
Verdict:
Solid mid-tier AI infra player, rising quickly, not yet a hyperscaler, but credible and growing.
D — Durability of Moat
Score: Developing
Existing moat sources:
- High switching costs for firms training large models.
- Regional regulatory moats (data residency requirements).
- Custom software layer atop raw compute.
- First-mover advantage in certain sovereign markets.
Missing moat pieces:
- No proprietary silicon like AWS (Graviton/Inferentia).
- No global multi-region dominance yet.
- Capex cycles can kill small infra players.
Verdict:
Early moat forming — but not locked. Need to watch if they create their own “flywheel”:
capacity → developer usage → higher utilization → cheaper costs → more capacity.
D — Downside Protection & Risks
Score: Weak → Moderate
Major risks:
- Capex trap: spending ahead of demand can kill cash flow.
- GPU supply risk: if NVIDIA deprioritizes them, they slow down.
- Commoditization: cloud compute races to the bottom long-term.
- Regulation: geopolitical shifts on data-center ownership.
- Margin pressure: hyperscalers can undercut pricing.
Balance:
- AI infra is the “railroads of the AI boom.”
- But railroads also went through brutal consolidation cycles — only a few survivors.
Verdict:
Not a widows-and-orphans stock. You must treat it like a growth compounder with real volatility baked in.
A — Alignment With Shareholder Value
Score: Moderate
Positives:
- Management speaking in unit economics terms: utilization, ROIC per rack, cost-to-compute.
- Growing with discipline (so far).
- Partnerships create network effects.
Negatives:
- Dilution risk typical of high-growth infrastructure companies.
- Large capex cycles may force equity raises.
- Still young → limited long-term track record.
Verdict:
Generally aligned, but not yet a proven capital allocator in downturns.
🟩 OVERALL GIDDA SCORE: 7.7 / 10
(High-growth, mid-risk, execution-dependent)
⚖️ Buffett-Style Summary
“If the AI boom stays hot, NBIS mints money from selling the shovels.
If the boom cools, heavy capex becomes an anchor.”
This is not a ‘sleep-well-at-night’ compounder yet.
It is a potential multibagger if it reaches hyperscaler scale without losing discipline.
🧠 NEW FRAMEWORK FOR THINKING ABOUT NBIS
The “Compute Sovereignty Trifecta”
A way to judge AI-infra companies going forward.
Every AI-infra business must dominate at least two of these:
- Hardware Sovereignty – control over GPUs / custom silicon / supply priority
- Data Sovereignty – ability to operate in strict jurisdictions (EU, MEA, India)
- Capital Sovereignty – ability to fund datacenter expansion cheaply
NBIS currently has:
- Good data sovereignty (positioning in strict markets)
- Decent hardware sovereignty (NVIDIA partnership)
- Weak capital sovereignty (high cost of capital vs hyperscalers)
For NBIS to re-rate upward, watch for signs of:
- Multi-billion funding rounds
- Proprietary hardware/software stack
- Expansion into sovereign-cloud contracts
- Rising utilization + falling unit costs consistently
This mental model gives you early indicators before the stock rerates.
Horizons 
Leave a comment