Secondary Marketplace for AI API Tokens/Credits

• Week 1-2: Landing page + Typeform intake for buyers/sellers + manual Slack/email matching • Week 3-4: Stripe integration for escrow + basic verification flow (screenshot upload, manual review) • Week 5-6: Document credential transfer process, send first 5 test transactions through, measure time-to-close • First dollar revenue achievable in 8 weeks if you manually source both sides (post in YC Slack, startup accelerator Discord servers) • Path to scale is ugly: each transaction requires manual verification, fraud checks, and coordination — labor-intensive until $500K+ GMV • Wedge strategy: focus on one provider (AWS or GCP credits from accelerator programs) where volume is highest and verification is simplest

• TAM estimate ~$500M/year in unused credits (11-17% of market claim), but addressable share is fraction of that • Primary supply: startup program credits (AWS $100K packages, OpenAI grants) — these are one-time windfalls, not recurring inventory • AI API market growing ($5K-50K/month typical startup spend in 2026), but most optimize usage before buying secondary credits • Active competitor AI Credits processed $20M cumulative (not annual), suggesting $5-10M annual GMV across 2+ years — implies sub-$50M SAM even with full category capture • Comparable AICreditMart and marketplace infrastructure exists, fragmenting already-small market • Most credits have 3-12 month expiration windows, compressing transaction velocity and creating perishable inventory problem

• Unit economics are broken at small scale: 10-15% marketplace take on $500 average transaction = $50-75 per deal, but CAC for buyers likely $100+ given niche audience • Inventory constraints kill velocity: perishable goods (3-12mo expiration) + episodic supply (only when startups over-provision) = inconsistent liquidity • Seller recovery at 40-70% means buyers save 30-60%, but direct startup programs offer $10K-150K free credits — free beats discounted • Existing player AI Credits at $20M total volume over 2+ years suggests $3-5M annual revenue at 15-20% take rate — barely venture-scale even as category leader • Legal risk depresses margins: need insurance, compliance checks, disputes — adds overhead that compresses thin marketplace margins • 90% repeat rate signals product-market fit for buyers, but seller side is one-time (can't sell credits twice) creating lopsided liquidity

• Real market exists: AI Credits (aicredits.co) already traded $20M+ with 200+ customers at 40-70% of face value, proving demand exists • Problem is real but narrow: $500M+ in AI credits expire unused annually, but most suppliers are startups with accelerator credits (Y Combinator, AWS Activate), not retail consumers • OpenAI explicitly prohibits credit transfers in Terms of Service (updated Jan 2026): "We prohibit and do not recognize any purported transfers, sales, gifts, or trades of Service Credits" with account termination as penalty • Enterprise buyers exist but prefer direct deals: 90% repeat buyer rate suggests sticky customers, but they're procurement teams buying $10K+ blocks, not individuals with unused $50 balances • Pain is episodic, not structural: companies over-provision once during fundraising, then adjust — not a recurring monthly problem worth solving

• Core tech is simple: marketplace matching + escrow + credential transfer (solved problem, 24-48hr verified transfers already standard) • No novel technology required: screenshots for verification, manual matching, Stripe for payments, basic CRM — standard marketplace stack • Biggest technical risk is verification fraud: preventing fake credit listings, stolen accounts, or post-transfer chargebacks • Integration challenge minimal: credits are account-level, transferred via email/password handoff or support tickets — no APIs to build • Could ship functional MVP in 4-6 weeks with no-code tools (Airtable + Typeform + manual matching) • Platform risk is high: reliant on providers not shutting down secondary market (OpenAI already has in ToS)