Valuation breakdowns — what builds the market cap, and what breaks it
For each company: price decomposition (cash vs. cash flow vs. growth optionality), customer concentration, growth-rate sensitivity to fair value, and decision trees for client wins and losses. Numbers are publicly disclosed where possible, modeled where required, and explicitly flagged either way.
Nvidia (NVDA)
Price decomposition (% of market cap)
2%
23%
46%
29%
2% net cash · 23% capitalized current FCF · 46% near-term growth · 29% terminal/optionality
Translation: 75% of equity value is the assumption that 2026–2032 FCF will be much larger than today's. The "current cash flow" anchor is only ~25% of the cap.
Customer concentration (data-center revenue)
Microsoft
~19%
Meta
~10%
Amazon
~7%
Google
~6%
Oracle
~5%
CoreWeave
~5%
Other
~48%
Top 4 ≈ 42% of data-center revenue. ~10–15% (CoreWeave + Nvidia-funded labs) is partly recycled from Nvidia's own equity stakes.
Sensitivity: fair value vs. revenue CAGR (next 5 years)
| WACC \ CAGR | 0% | 10% | 20% | 30% (consensus) | 40% |
|---|---|---|---|---|---|
| 7% | $80 | $120 | $170 | $240 | $340 |
| 9% | $65 | $95 | $135 | $185 | $255 |
| 11% | $55 | $80 | $110 | $150 | $200 |
Spot ≈ $180 (Apr 2026) · Base case $185 ≈ +3% · Bear ($55) −69% · Bull ($340) +89%
Decision tree (next 12 months)
Hyperscalers expand FY26 capex 25%+ P ≈ 35% +25% to +40%
+ China export-control loosening P ≈ 15% +15% additional
Status quo — 30% revenue growth lands P ≈ 30% ±10%
One hyperscaler pauses (Meta or Google example) P ≈ 20% −20% to −30%
Two hyperscalers pause (capex committee revolt) P ≈ 10% −40% to −55%
CoreWeave default + funding-round failure P ≈ 5% −55% to −70%
Probability-weighted expected return ≈ +5% over 12 months. Distribution skew: long upside tail capped by current valuation; downside tail open if multiple hyperscalers pause.
Microsoft (MSFT)
Price decomposition
2%
53%
30%
15%
53% capitalized current FCF (Office, Windows, Server) · 30% Azure/AI growth · 15% OpenAI stake + terminal
Translation: Microsoft is more anchored to today's cash flow than Nvidia. The mature franchises (Office, Windows, Xbox, on-prem Server) cover ~half the market cap.
Revenue concentration by segment
Productivity & Business Processes
32%
Intelligent Cloud (Azure)
43%
More Personal Computing
25%
Within Azure: ~25–30% of growth attributed to OpenAI / AI services in latest reported quarter. Single largest customer (OpenAI) ≈ 6–8% of total Azure revenue.
Sensitivity: fair value vs. Azure CAGR
| WACC \ Azure CAGR | 10% | 20% (consensus) | 30% |
|---|---|---|---|
| 7% | $420 | $520 | $640 |
| 9% | $380 | $465 | $565 |
| 11% | $345 | $420 | $510 |
Spot ≈ $465 (Apr 2026) · Base case $465 ≈ flat · Bear ($345) −26% · Bull ($640) +38%
Decision tree
Copilot ARR > $20B run rate by EOY P ≈ 25% +15% to +25%
Status quo — Azure 20%, Office strong P ≈ 45% ±5%
OpenAI write-down ($5B+ impairment) P ≈ 15% −10% to −15%
Azure decel below 18% P ≈ 10% −15% to −25%
Antitrust / OpenAI partnership disruption P ≈ 5% −20% to −30%
AMD (AMD)
Price decomposition
2%
18%
55%
25%
20% anchored to current cash flow · 80% on AI growth that depends on OpenAI execution
Customer concentration in MI300/MI400 ramp
OpenAI commitment
~55%
Meta
~15%
Microsoft
~12%
Oracle / others
~18%
OpenAI is the single biggest swing factor in AMD's growth story. They also got 10% AMD warrants — circular reflexivity.
Sensitivity: fair value vs. AI revenue scenarios
| Scenario | FY26 AI rev | FY28 AI rev | Fair value |
|---|---|---|---|
| Bear: OpenAI defers | $3B | $5B | $80 |
| Base: deal as announced | $8B | $22B | $150 |
| Bull: deal + 2nd hyperscaler win | $12B | $35B | $240 |
Spot ≈ $150 (Apr 2026) · Base case $150 ≈ flat · Bear ($80) −47% · Bull ($240) +60%
Decision tree (next 18 months)
MI400 wins second hyperscaler (Google or Amazon) P ≈ 20% +50% to +80%
OpenAI ramp on schedule P ≈ 35% ±15%
OpenAI defers MI400 by 2 quarters P ≈ 25% −25% to −35%
OpenAI cancels MI400 (Nvidia exclusivity push) P ≈ 15% −45% to −60%
Capex pause + warrants OOM P ≈ 5% −60% to −70%
Oracle (ORCL)
Price decomposition
35%
50%
15%
~50% of equity value depends on the $300B OpenAI deal performing as announced.
Customer concentration in OCI growth
OpenAI
~60%
TikTok / ByteDance
~8%
xAI
~5%
Other enterprise
~27%
Single counterparty controls more than half of forward OCI revenue. Most extreme concentration of any mega-cap on this list.
Sensitivity: fair value vs. OpenAI deal performance
| Scenario | OCI 5yr revenue | Fair value |
|---|---|---|
| Bear: OpenAI renegotiates -50% | $150B | $130 |
| Base: deal as announced | $300B | $240 |
| Bull: deal + 2nd anchor (e.g. xAI) | $420B | $340 |
Spot ≈ $240 (Apr 2026) · Base case $240 ≈ flat · Bear ($130) −46% · Bull ($340) +42%
Decision tree
Add second $50B+ anchor (xAI scale-up) P ≈ 15% +30% to +45%
OpenAI deal lands without major slip P ≈ 45% ±10%
OpenAI defers $50B of build P ≈ 25% −20% to −30%
OpenAI restructures, deal repriced -40% P ≈ 12% −40% to −55%
Counterparty crisis + Oracle debt downgrade P ≈ 3% −60%+
CoreWeave (CRWV)
Price decomposition
75%
25%
100% of equity value is the contracted backlog ($30B+ remaining performance obligations) + the bet that more customers materialize. Zero anchoring to current cash flow.
Customer concentration
Microsoft
~62%
OpenAI
~15%
Meta
~7%
Other
~16%
77% of revenue from two customers. Highest concentration in the public AI infrastructure stack.
Sensitivity: per-share fair value vs. customer scenarios
| Scenario | FY26 utilization | Equity value | $/share (≈500M sh) |
|---|---|---|---|
| Bear: MS reduces takeup -25% | 65% | $8B | $16 |
| Base: contracts perform | 85% | $40B | $80 |
| Bull: Anthropic + xAI added | 95% | $70B | $140 |
Spot ≈ $80 (Apr 2026) · Base $80 ≈ flat · Bear $16 −80% · Bull $140 +75%
Decision tree
Anthropic or xAI signs $5B+ multi-year P ≈ 20% +40% to +70%
Existing contracts perform on schedule P ≈ 40% ±15%
MS reduces incremental takeup P ≈ 20% −35% to −50%
OpenAI delays / cancels CoreWeave use P ≈ 12% −50% to −70%
Debt covenant trip + restructuring P ≈ 8% −85% to −100%
Probability-weighted fair value
| Branch | P | Move (midpoint) | $/share | Contribution |
|---|---|---|---|---|
| Anthropic / xAI signs | 20% | +55% | $124 | $24.80 |
| Status quo | 40% | ±0% | $80 | $32.00 |
| MS reduces takeup | 20% | −42% | $46 | $9.20 |
| OpenAI delays | 12% | −60% | $32 | $3.84 |
| Debt restructuring | 8% | −92% | $6 | $0.48 |
| E[V] | 100% | −12% | $70 | $70.32 |
Probability-weighted fair value ≈ $70 / share vs spot $80 → −12%.
But the standard deviation across branches is $42: outcomes range from $6 to $140.
The expected value undersells the actual risk profile here — this is a bimodal distribution,
not a Gaussian. Position-sizing for CoreWeave should treat it more like a binary bet
(lottery shape) than a typical equity (continuous Gaussian).
Tesla (TSLA)
Price decomposition
3%
15%
25%
57%
~57% of equity value is contingent narrative (FSD, Robotaxi, Optimus, xAI) — none materially revenue-generating today.
Revenue mix today vs. priced future
Auto sales (today)
80% rev
Energy (today)
8% rev
Services (today)
12% rev
FSD/Robotaxi (priced)
~55% mcap
The valuation gap between current revenue mix and what the market is pricing is the largest of any mega-cap.
Sensitivity: fair value vs. FSD/Robotaxi materialization
| Scenario | Robotaxi FY28 rev | Fair value/share |
|---|---|---|
| Bear: FSD plateau, no Robotaxi | $0 | $110 |
| Base: Limited Robotaxi rollout | $5B | $280 |
| Bull: National Robotaxi + Optimus pilot | $30B | $520 |
Spot ≈ $280 (Apr 2026) · Base case $280 ≈ flat · Bear ($110) −61% · Bull ($520) +86%
Decision tree
Robotaxi launches in 5+ cities by EOY P ≈ 15% +25% to +50%
Auto demand rebounds, FSD take-rate > 25% P ≈ 25% +15% to +25%
Status quo P ≈ 25% ±10%
Auto margins compress further P ≈ 25% −20% to −35%
Robotaxi rollback / regulatory hit P ≈ 10% −40% to −60%
Read this carefully
- Sensitivity tables are not predictions. They show how fair value moves under different growth and discount-rate inputs — so you can see how much of today's price is bet on which assumption.
- Decision-tree probabilities are subjective. They're modeled to be coherent (sum to 1) and reflect a plausible rather than canonical view. Adjust the priors mentally as you read.
- Customer concentration data for hyperscalers is approximate and reconstructed from 10-K segment disclosures + industry estimates. Companies don't disclose customer percentages by name. Treat ranges, not point values.
- Price decomposition is a DCF backsolve: given today's price and a discount rate, what implicit growth profile is required? The split into "current FCF / growth / optionality" is a view, not an audited number.
- The point isn't to get a "fair value" right. It's to make explicit what assumption(s) the current price requires, and what magnitude of revaluation each scenario branch implies.