02 UNIT ECONOMICS / 36-MONTH DCF

Project Economics

Test durable cash flow, not just a promising idea. Put MAU, tokens, conversion, sponsorship, and cost into one explicit model.

OUTPUT: Token cost · BEP · NPV
HORIZON: 36 months
PRICES: OpenAI · 2026-07-17

ASSUMPTIONSBase operating case

01 / Audience & usage

MAUusersSessions / MAU/moAI calls / sessioncallsAI feature adoption%

02 / Inference

MODEL / STANDARD PRICE

INPUT $0.2/1MCACHE READ $0.02/1MCACHE WRITE $0.20/1MOUTPUT $1.25/1M

Official source: GPT-5.4 nano pricing · checked 2026-07-17

Input / calltokensOutput / calltokensCache-read input share%Cache-write input share%Batch/Flex share%Web search / call%FX assumption₩/$

03 / Revenue

Paid conversion%Subscription₩/moPayment loss%Ads/affiliate ARPU₩/MAUSponsor MRR₩B2B MRR₩

04 / Cost & finance

Other variable cost₩/MAUFixed operating cost₩/moInitial build cost₩Monthly MAU growth%Annual hurdle rate%

LIVE MODEL / 36 MONTHSNPV > 0

PORTABLE EXPERIMENT · V1.vlsilab.jsonLOCAL FILE · NO UPLOAD

Recent

Carries model choice and bounded planning assumptions only; computed results are regenerated in this browser.

36-MONTH NET PRESENT VALUE₩29.1M

Initial build ₩10M · annual hurdle 25% · zero terminal value

AI COST / MONTH₩17,28710,500 calls

COST / REQUEST$0.001176tokens + expected search

BREAK-EVEN MAU1,733minimum at current monetization

GROSS MARGIN95.4%before fixed cost

01 / CURRENT MONTH

From revenue to cash flow

Paid subscription₩960,300

Ads / affiliate₩600,000

Sponsor + B2B₩1,000,000

AI + other variable− ₩117,287

Fixed operating cost− ₩1,500,000

Monthly cash flow₩943,013

02 / SENSITIVITY

Does the case survive at half strength?

DOWNSIDE-₩17.2M50% MAU and sponsor; 0% growth

BASE₩29.1MCurrent inputs

UPSIDE₩70.3M150% MAU; growth +1 pp

If NPV is positive only in the base case, secure a paid partner or sponsor—or lower fixed cost—before launch. This planning screen is not a forecast or investment advice.

03 / COST LOGIC

What creates token cost?

C/request = uncached input + cache read + cache write + output + expected tool calls
AI/month = MAU × adoption × sessions × calls × C/request
NPV = −initial cost + Σ monthly cash flow / (1 + monthly hurdle rate)^m

Official price snapshot checked July 17, 2026. Batch/Flex discounts only the token portion by 50%; expected web-search calls remain undiscounted. GPT-5.6 Luna cache writes cost 1.25× uncached input, and prompts above 272K input tokens price the full request at 2× input and 1.5× output. Verify current values on the official model pricing page ↗ before relying on the model.