OpenAI and Anthropic Revenue Breakdown

Original Article ↗ Hacker News Discussion ↗

Valuation, Revenue & Losses

  • OpenAI reportedly has ~$3.6–3.7B revenue (mostly $20/mo subscriptions) but is expected to lose ~$5B+ this year; some say real loss with stock comp could be $8–10B.
  • Several commenters note the original “P/E ~43” claim is incorrect because earnings are negative; actual P/E is undefined/negative.
  • Some compare price-to-sales (~40x+) to big tech (e.g., Amazon ~3x), calling the valuation “peanuts vs. price”; others say high losses are typical for hypergrowth.

Business Model, Unit Economics & Churn

  • Roughly 75% of revenue is estimated from ChatGPT subscriptions, ~25% from API/enterprise (including Microsoft-related usage).
  • Debate on whether subscriptions or API are more profitable; one estimate claims API has ~50% gross margins and that unlimited $20 plans are loss-making for heavy users.
  • Several note high free usage (≈180M users) vs. ~11M paying users and significant churn after 1–3 months.
  • Some expect eventual “closing of the hand”: worse free offering and/or higher prices to force upgrades.

Moat, Competition & Commoditization

  • Strong brand (“ChatGPT” as generic LLM term) and UX seen as key advantages; others argue models are commodity and switching providers is trivial (e.g., via Bedrock).
  • Concerns that open-source and rival models (Anthropic, Google, Meta, Chinese labs, Pika, etc.) erode proprietary moats.
  • Debate over whether this ends as a commodity, low-margin “airline/WeWork” situation vs. a dominant, ad-funded or subscription giant akin to Google Search.

Microsoft Relationship & Infrastructure Risk

  • Discussion of complex IP/profit-sharing deal; some worry Microsoft is racing to replace OpenAI and that OpenAI would struggle to fund its own $100B+ datacenter buildout.
  • Others think OpenAI’s valuation, brand, and access to capital alleviate this risk.

Adoption, Use Cases & Labor Impacts

  • Many pay for AI tools and find them indispensable, especially for coding; others find free versions or local models “good enough” and refuse subscriptions.
  • Split views on long-term impact on developers: from “just another productivity tool” to serious threat to junior roles and high salaries.

Investment Angles

  • Suggestions to invest indirectly via GPUs, datacenters, power generation, or “AI shovel sellers,” rather than frontier model labs themselves.