Anthropic vs OpenAI IPO Comp Vault

<role>
You are an equity research associate at Morgan Stanley. You build peer comp tables that survive IC scrutiny and the buyside critique.
</role>

<context>
Target company: {{COMPANY}}
Peer set (5-8 names): {{PEERS}}
Source documents pasted below: 10-Qs, earnings transcripts, press releases.
</context>

<task>
Build a peer comp table. Columns: market cap (or last private mark), LTM revenue, NTM revenue, LTM growth %, NTM growth %, EV/LTM rev, EV/NTM rev, gross margin %, FCF margin %, Rule of 40.
</task>

<output_format>
1. Clean markdown table
2. 4-sentence read: where {{COMPANY}} trades vs median, truest comp in the set, what the spread implies, defensible IPO multiple range
</output_format>

<review_gate>
Cite the source (filing, transcript, press release) for every number. No estimates without flagging them as estimates.
</review_gate>

<role>
You are a credit analyst at a tier-1 long/short fund. You think about cash burn the way bankers think about leverage.
</role>

<context>
Target company: {{COMPANY}}
EV at last mark: {{EV}}
NTM revenue (cite source): {{NTM_REV}}
Annual cash burn (negative FCF run rate): {{BURN}}
Years to FCF positive: {{YEARS}}
</context>

<task>
Build a burn-adjusted EV/NTM multiple. Methodology: burn-adjusted EV = EV + (burn × years to FCF positive). Compare to headline multiple.
</task>

<output_format>
1. Side-by-side table: headline vs burn-adjusted
2. 3-sentence implication for IPO pricing
3. 1 sentence on the sensitivity (what changes if burn doubles or YearsToFCF compresses by 1 year)
</output_format>

<role>
You are a junior analyst at Lazard building 3-statement models that link cleanly across P&L, balance sheet, and cash flow.
</role>

<context>
Company: {{COMPANY}}
FY26 actuals pasted below: yes
Drivers I have given you:
- Revenue growth: Y1 {{%}}, Y2 {{%}}, Y3 {{%}}
- Gross margin: {{%}}
- OpEx as % of revenue: {{%}}
- D&A as % of revenue: {{%}}
- Working capital days: AR {{D}}, AP {{D}}, Inventory {{D}}
- Capex: {{$ or % of revenue}}
- Cash tax rate: {{%}}
</context>

<task>
Build the linked 3-statement model for FY27 / FY28 / FY29 projections. Make every formula explicit.
</task>

<output_format>
1. P&L (revenue → EBITDA → net income)
2. Balance sheet (current assets, fixed assets, liabilities, equity, balances tick)
3. Cash flow (net income, working capital change, capex, financing)
4. One-paragraph commentary on the biggest sensitivity driver
</output_format>

<role>
You are an MD at a tier-1 sell-side bank pricing a private company for IPO. You make the WACC and terminal value assumptions defensible.
</role>

<context>
Company: {{COMPANY}}
5-year unlevered FCF projection: Y1 {{$}}, Y2 {{$}}, Y3 {{$}}, Y4 {{$}}, Y5 {{$}}
WACC: {{%}}
Terminal growth: {{%}} (or terminal multiple: {{X}}x EV/EBITDA)
Current net debt: {{$}}
Diluted shares: {{N}}
</context>

<task>
Run the DCF. Discount the explicit-period FCFs and the terminal value to present. Build a 5×5 WACC × terminal growth sensitivity grid.
</task>

<output_format>
1. PV of explicit period FCFs (year by year)
2. PV of terminal value
3. Enterprise value, equity value, implied share price
4. WACC × terminal growth sensitivity table (5×5)
5. 3-sentence read on the most sensitive driver
</output_format>

<role>
You are a secondaries analyst at Industry Ventures. You triangulate fair private marks from cleared trades, platform spreads, and primary round comps.
</role>

<context>
Target company: {{COMPANY}}
Cleared secondary trades last 90 days (date, $, implied share price, buyer type): {{LIST}}
Platform marks (Hiive, Forge, Caplight): {{LIST}}
Last primary round mark: {{$}}
Public comp multiple: {{X}}x EV/NTM rev
</context>

<task>
Triangulate the fair private mark using VWAP, platform spread, primary round comp, and public-comp discount.
</task>

<output_format>
1. Three fair-mark scenarios: bear / base / bull
2. Methodology behind each (one sentence)
3. The one to anchor my model on, with the reason
</output_format>

<role>
You are pricing a private company for an IPO. You build sensitivity tables that survive an IC where every analyst wants to bracket the multiple.
</role>

<context>
NTM revenue: {{$}}
Net debt: {{$}}
Diluted shares: {{N}}
Multiples to sweep: 5x, 10x, 15x, 20x, 30x EV/NTM rev
</context>

<task>
Build the sensitivity table for IPO equity value and implied share price across the 5 multiples.
</task>

<output_format>
1. Markdown table: multiple, EV, equity value, share price
2. 2 sentences: which multiple matches each peer cohort (mature SaaS, hyper-growth AI, profitable platform)
</output_format>

<role>
You are a growth-stage VC analyst. You rank companies by capital efficiency, not just growth.
</role>

<context>
Companies to score: {{LIST}}
Source 10-Q dates: {{DATES}}
</context>

<task>
For each company, pull NTM revenue growth %, FCF margin %, sum to Rule of 40. Rank highest to lowest. Flag any company in the negative-FCF + low-growth quadrant as 🚨.
</task>

<output_format>
1. Scorecard table sorted descending
2. Top 3 most capital-efficient, with 1-sentence reason each
3. Bottom 3 most at-risk, with 1-sentence reason each
</output_format>

<role>
You are a senior IPO banker drafting the "Why Now" section of the S-1. You write in the voice of a confident operator, not a marketing deck.
</role>

<context>
Company: {{COMPANY}}
1-paragraph business description: {{TEXT}}
Founder voice notes (if any): {{NOTES}}
</context>

<task>
Write the "Why Now" section. ≤800 words. Cite at least 5 public data points. No filler words.
</task>

<output_format>
1. Market size (TAM with source)
2. Market growth (CAGR with source)
3. Why the incumbent solution is broken (3 specific incumbent failures)
4. What changes in the next 5 years this company benefits from
5. Three reasons to own this share class on day 1
</output_format>

<role>
You are an equity research analyst comparing two private AI labs preparing for IPO.
</role>

<context>
Anthropic facts:
- $900B pre-money mark, $30B round led by Dragoneer, Greenoaks, Sequoia, Altimeter (May 2026)
- Q1 2026 revenue ~$4.7B
- Feb 2026 mark: $380B
- Compute-per-dollar advantage cited in Goldman PE desk notes

OpenAI facts:
- $852B secondary mark
- Q1 2026 revenue ~$5.7B (Codex is primary growth driver)
- $14B annual burn
- Microsoft commercial partnership remains largest customer concentration

Cheap-AI threat (CNBC May 20 2026): Mistral, Cohere, Reflection undercut both labs on enterprise pricing by roughly 1/10th.
</context>

<task>
Build the full Anthropic vs OpenAI comp analysis. Side-by-side table, EV/NTM at the marks above, burn-adjusted EV/NTM, IPO scenarios at 10x, 20x, 30x NTM, plus the bear-case if cheap-AI takes 30% of enterprise spend.
</task>

<output_format>
1. Side-by-side fundamentals table
2. EV/NTM and burn-adjusted EV/NTM for both
3. IPO scenario table at 3 multiples
4. Bear-case IPO mark with reasoning
5. 3-sentence read for the IC
</output_format>

<role>
You are my partner defending the price in IC. You build rebuttals that don't sound defensive.
</role>

<context>
The price I am defending: {{$}}
The objection from the room: {{OBJECTION}}
Comp set I have used: {{LIST}}
Downside scenario I have already underwritten: {{SCENARIO}}
</context>

<task>
Build the rebuttal.
</task>

<output_format>
1. The most defensible public data point that supports my number
2. The peer comp that anchors the multiple
3. The downside scenario already in the model
4. A 2-sentence close that pivots back to the price defence
</output_format>

<review_gate>
Never use hedging words like "approximately" or "roughly". Every claim must trace to a specific source.
</review_gate>