How will OpenAI compete?
TLDR
- OpenAI has ~900M weekly users and no durable competitive lead: models are matched, engagement is shallow, and incumbents have distribution.
Key Takeaways
- Roughly half a dozen organizations ship competitive frontier models with equivalent capabilities; no known mechanism exists for one to achieve a permanent lead.
- 80% of ChatGPT users sent fewer than 1,000 messages in all of 2025, averaging under three prompts per day; only 5% pay.
- Google and Meta are gaining market share by leveraging existing distribution against a product that typical users cannot meaningfully distinguish from ChatGPT.
- OpenAI’s product heads do not control the roadmap: labs surface breakthroughs and product teams are tasked with turning them into UI, not setting strategy.
- Sam Altman’s stated platform vision frames capex as a virtuous circle and positions OpenAI as a full-stack platform, but the CFO’s own diagram shows a 1:1 capex-to-revenue relationship, not a flywheel.
Why It Matters
- OpenAI’s “capability gap” framing, acknowledging a gap between what models can do and what users actually do, is an admission that product-market fit is not yet established.
- If the next wave of AI value comes from new experiences built on top of raw models, OpenAI has no clear structural reason to capture that over thousands of competing builders.
- Anthropic holds benchmark leads but has no consumer strategy; ChatGPT has scale but thin engagement; neither position is clearly durable against Google and Meta’s distribution.
Benedict Evans · ** · Read the original