P2B-C3 · Valuation/Earnings Terminology Dictionary¶

Key Insight

Translate terms when you see them — don't bypass them. Bypassed terms will bite you.

Layer 2 · Analyzing a Single Stock — After learning the industry, now drill down to one stock

L2-C3 (5 chapters total). After this chapter, you'll be able to first classify any earnings term into one of the 4 dimensions, then read the actual numbers.

1. The Problem: Bypassed Terms Will Bite You¶

You're writing your C2 yaml and encounter "ORCL RPO $553B."

You don't know what RPO is → you don't include it in supports
But ORCL Q3 earnings +30% was entirely driven by RPO upward revision → your thesis misses the core driver
6 months later, when OpenAI RPO concentration blows up, you won't see it coming (because you never classified that number)

Bypassing 1 term = your thesis misses at least 1 support or 1 red_flag. Compound that over 10 terms, and your thesis is no better than rolling dice.

2. The Solution: Classify Every Term into 4 Dimensions First, Then Read the Numbers¶

Don't treat the glossary as a dictionary to memorize (you'll forget it). The method:

Encounter a term → First ask: Which dimension is this — WHAT / WHY / SO WHAT / RISKS?
After classification, fill it into the corresponding yaml field (support / catalyst / price_outlook / red_flag)
Read the actual numbers — calibrate your sense of "high / low / healthy" using real company cases like NVDA / TSM / MSFT

All terms below are grouped by the 4 dimensions, not by textbook categories like "Revenue / EPS / Valuation."

3. How It Works: A 4-Dimensional Terminology Library¶

3.1 WHAT Dimension — What is the current narrative? (→ filling `supports`)¶

Segment Revenue¶

Large companies have more than one business. Segment revenue tells you where the money comes from.

NVDA: Data Center (~88%), Gaming (~8%), Pro Viz (~2%), Auto (~2%)
MSFT: Productivity (Office) / Intelligent Cloud (Azure) / More Personal Computing
AMZN: AWS / Online Stores / 3P Seller / Subscription / Advertising

⚠️ Key Insight: 30% revenue growth looks good, but if Data Center is up +75% while other segments are declining, that's "AI pulling + other segments deteriorating." Only by looking at segment breakdown do you know the thesis strength.

Backlog + RPO (Remaining Performance Obligations)¶

The amount of signed contracts where revenue hasn't been recognized yet — the core forward-looking metric in the AI era.

Company	Number	Meaning
ORCL Q3 FY26 RPO	$553B	~$300B is OpenAI contracts (single customer 54%!)
MSFT RPO	$260B	cloud commitments
CRWV backlog	~$99.4B	NVDA $36.6B stake + Neocloud lock-ins
NVDA	Not directly disclosed	But inferable from hyperscaler 12-month lock-ins

⚠️ ORCL example: The total RPO number is impressive, but single-customer concentration = core thesis risk (this is also a RISKS dimension item — see 3.4).

Capex Cycle 2^nd Derivative (hyperscaler capex guidance changes)¶

Look at whether hyperscalers raise or lower their next-year capex guidance:

Guidance Direction	Signal
Raised	Entire AI chain is bullish (NVDA / COHR / CEG)
Flat	AI narrative plateaus, 2^nd derivative = 0
Lowered	Major bear catalyst

Key: Look at the 2^nd derivative, not just the absolute value. MSFT is already at $80B/yr; going to $85B = 5% increase, not a big positive. Going from $80B → $100B (25% increase) is a bull catalyst.

Wafer Capacity / HBM Capacity¶

Wafer foundry capacity (TSM 3nm / Samsung 2nm) and HBM memory capacity (SK Hynix / Micron / Samsung) are real physical bottlenecks:

TSM 3nm capacity determines the shipment ceiling for NVDA / AMD / Apple
HBM shortage directly caps NVDA H200 shipments

Hyperscaler vs Neocloud¶

Type	Representative	Characteristics
Hyperscaler	MSFT Azure / GOOGL Cloud / AMZN AWS	Large multi-business clouds, diversified customers
Neocloud	CRWV (CoreWeave) / NBIS (Nebius) / ORCL OCI	AI-dedicated, highly concentrated customers

Difference: Neocloud has high customer concentration (CRWV: MSFT 60%+, ORCL: OpenAI 54%) → explosive growth but high customer concentration risk (also a RISKS dimension input).

3.2 WHY Dimension — What are you more confident about than the market? (→ filling `core_thesis` + `confidence`)¶

Pre-announcement / Guidance¶

Pre-announcement: Key numbers disclosed early (usually 2 weeks) before earnings
Guidance: Next quarter / full-year expectations disclosed on earnings day

⚠️ The market often reacts to guidance, not the current quarter's result. If your WHY can anticipate the guidance direction (e.g., "MSFT FY27 capex guide will surprise upside because of OpenAI commit"), you're 1 step ahead of the market.

Smart Money Divergence¶

When Bridgewater sells but Citadel buys the same stock — this indicates internal disagreement among smart money. A single 13F signal < average expectation.

If your WHY can explain the divergence (e.g., Bridge sees macro risk, Citadel sees stock-specific catalyst), you're more granular than someone just looking at the 13F total.

3.3 SO WHAT Dimension — Price Impact (→ filling `price_outlook` + `catalysts`)¶

Revenue — The Starting Number¶

How much the company sold this quarter (before costs):

NVDA Q4 FY26: $39.3B
TSM Q1 FY26: $35.9B
MSFT Q1 FY26: $82.9B

Look at YoY growth and QoQ growth. For AI companies, YoY +50%+ is "healthy," <30% is a "slowdown warning."

Gross Margin¶

(Revenue - direct production costs) / Revenue %:

NVDA: 75% (software + premium chip pricing power)
TSM: 53-58%
MU: 35-45% (memory cyclicality)
HPE: 30% (servers, low margin)

Look at trend — rising means improving pricing power (e.g., NVDA H200 vs H100), falling means competition or cost issues.

Operating Margin¶

(Revenue - all costs, including R&D, SG&A) / Revenue %:

MSFT: 45% (cloud scale effects)
NVDA: 60%+ (jaw-dropping)
TSM: 47%
AMD: 7-12%

Net income / shares outstanding:

NVDA Q4 FY26 non-GAAP EPS: $0.89 (split-adjusted)
AMD Q4 2025 non-GAAP EPS: $0.85

(GAAP vs non-GAAP difference is a RISKS dimension item — see 3.4)

FCF (Free Cash Flow)¶

Operating cash flow - capex. The money that actually lands in the company's pocket:

MSFT TTM FCF: $74B
GOOGL TTM FCF: $72B
NVDA TTM FCF: $60B
ORCL TTM FCF: -$24.7B (negative!) ← AI capex far exceeds operating cash flow

Capex (Capital Expenditure)¶

Money spent on long-term assets:

MSFT FY26 capex expectation: ~$80B
GOOGL FY26 capex expectation: ~$75B
AMZN FY26 capex expectation: ~$90B
META FY26 capex expectation: ~$60B

Total = $300B+, almost entirely flowing to NVIDIA / TSMC / Micron / Coherent / Vertiv / Power — that's why all AI picks-and-shovels stocks are up.

P/E / PEG / EV/Revenue / Rule of 40¶

Multiple	How to Read	Example
fwd P/E	Stock price / next year's expected EPS	NVDA 30-35x · MSFT 30x · MU 8-10x
PEG	P/E ÷ EPS growth rate (%). Entry-level rule: < 1 cheap, > 2 expensive	MU 0.4 · NVDA 1.0 · LRCX 2.0 · AVGO 0.9
EV/Revenue	Suitable for unprofitable / low-margin (SaaS / Neocloud)	CRWV 12-15x · NBIS 10x · SNOW 13x · CBRS 77x
Rule of 40	Growth rate + operating margin ≥ 40% = healthy growth stock	CRWD 60% ✓ · DDOG 50% ✓ · MDB 29% ✗

Fiscal Year — Don't Mix Up Calendar¶

Company	FY27 Range
NVDA	2026-02 → 2027-01 (similar to Apple)
MSFT	2026-07 → 2027-06
ORCL FY26	2025-06 → 2026-05
CSCO FY26	2025-08 → 2026-07

⚠️ When you see "Q1 FY27," always verify which calendar month it corresponds to. A wrong catalyst date = thesis verification in the wrong quarter.

3.4 RISKS Dimension — What Would Trigger Your Exit? (→ filling `red_flags` + `trigger`)¶

GAAP vs non-GAAP Gap¶

GAAP includes "non-cash" costs like stock-based compensation; non-GAAP excludes them. The market focuses more on non-GAAP, but:

⚠️ A GAAP - non-GAAP gap that's too large (e.g., >20%) is a yellow flag — the company may be masking weak operations with incentives.

FCF Turning Negative (AI Era Specific)¶

Many AI companies' FCF turns negative due to massive capex (MSFT / AMZN / ORCL are all close to or already negative).

⚠️ This is a thesis-critical question: Is the capex for the future (will earn it back) or a black hole (never fills up)? If you can't answer → red_flag.

Trigger format: "If FCF is negative for 4 consecutive quarters AND revenue growth rate drops below 20% → trigger"

Customer Concentration (Neocloud / RPO Source)¶

CRWV: MSFT 60%+
ORCL OCI: OpenAI 54% (RPO $300B / $553B)
CRWV: NVDA $36B stake = circular financing concern

⚠️ Single customer >50% = red_flag. Trigger: "If the large customer discloses order reduction OR the large customer's own fundamentals deteriorate → trigger"

13F Lag + Limitations¶

13F-HR: Institutional (>$100M AUM) quarterly disclosure of holdings.

⚠️ Limitations:

6-week lag (holdings disclosed on 5/15 are from 3/31)
Does not show short positions
Does not show options (Aschenbrenner's $1.57B NVDA put is a rare 13F case)

Before using "13F increased holdings" as a support in your thesis, ask: Has the macro / catalyst changed in those 6 weeks? If yes, the support is invalid.

Capex 2^nd Derivative = 0 (RISKS Perspective)¶

Section 3.1 covered this from the WHAT dimension (catalyst). But the same data from the RISKS dimension:

Once hyperscaler capex plateaus → the entire AI chain de-rates
Trigger: "MSFT or GOOGL FY27 capex guide < 10% YoY growth"

4. vs. What You Already Know from C2¶

C2 gave you the yaml skeleton — but if you don't understand the words you're putting into the skeleton (RPO / FCF / fwd PE / Rule of 40), you'll put them in the wrong place:

Putting RPO into supports seems OK, but in the ORCL example, RPO is also a RISK (single-customer concentration) — if you only put it in 1 place, you miss 1 place
Putting "Capex raised" into catalysts_90d is correct, but "Capex flat" is also a RISKS trigger condition
Putting "13F increased holdings" into supports seems strong, but you don't know the 6-week lag caveat — the market may have already changed in those 6 weeks

C3 teaches you to first classify each term into one of the 4 dimensions, then fill it in. A single term may fall into multiple dimensions simultaneously (RPO is both WHAT and RISKS), and you need to record it in all relevant yaml fields.

5. Try It: Fill in 1 Real Number for Each of the 4 Dimensions for Your Ticker¶

Task: Using the ticker you selected in C1/C2, fill in at least 1 term with a number and a source for each dimension (~30 minutes)

Dimension	Your Ticker Example	Source (Earnings / 13F / Which Transcript)
WHAT	___ (segment / RPO / backlog / capex cycle)	___
WHY	___ (guidance direction / smart money divergence)	___
SO WHAT	___ (PE / PEG / FCF / Rule of 40)	___
RISKS	___ (concentration / GAAP gap / capex 2^nd derivative)	___

Self-check:

Each cell has a specific number (not adjectives like "high" / "low")
Each cell has a source (company earnings PDF / Motley Fool transcript / 13F filing URL)
You can explain to a friend whether each number is high or low relative to the industry benchmark (e.g., NVDA 75% gross margin is higher than TSM 55% because of software + pricing power)

All 4 cells filled + all self-checks yes → your C2 yaml now has real numbers for supports/red_flags/catalysts.

6. What's Next¶

You know the terms, and you've filled in the yaml with numbers. But you haven't run through a complete stock analysis yet:

How do you connect the transcript to the 4-dimensional yaml?
What are KPI predictions? How do you verify them?
How do you update the yaml after earnings?

→ L2-C4 · Real Walkthrough (NVDA) uses the NVDA Q4 FY26 → Q1 FY27 real case to walk through the full 6-step cycle.

7. Deep Dive (optional): GAAP vs non-GAAP / FCF Turning Negative in the AI Era¶

Click to see 3 AI-era term traps

Trap 1: non-GAAP beautification — but the gap also tells you something

NVDA non-GAAP EPS $0.89, GAAP EPS $0.81 — gap ~10%, normal. Some SaaS GAAP-non-GAAP gap >40% → 50% of the company's profit is an accounting concept, not cash. The gap itself is a signal.

Trap 2: FCF turning negative — capital expenditure vs. black hole

MSFT / GOOGL / META FCF is still positive but narrowing rapidly. ORCL has already turned negative. The question isn't "negative" itself, but the capex / revenue ratio:

Healthy: capex / revenue ~15-20%
Warning: ~30%+
ORCL FY26: close to 40% (due to OpenAI commit)

Trigger format: "If capex/revenue > 35% for 2 consecutive quarters AND revenue growth rate <30% → ORCL RISKS triggered"

Trap 3: 13F signal decay in the AI era

Many AI hedge funds use options + shorts to express their view (e.g., Aschenbrenner's $1.57B NVDA put). 13F only shows long stock = you systematically miss sophisticated players' true positioning.

Mitigation: When looking at 13F, also check for abnormal options open interest (especially large-strike long-dated puts/calls).