What happened
OpenAI has revised its annualised revenue projections downwards by $20 billion. TechCrunch reports that the company told investors its revenue is approaching $50 billion, not the $70 billion figure reported just over a week ago.
The previous, higher figure was reportedly an attempt by investors to create a direct comparison with Anthropic's run rate. OpenAI and Anthropic calculate their revenue differently — OpenAI does not include sales made by its cloud partners, while Anthropic does.
How the room's reading it
The revision is fuelling chatter about the true cost of building frontier models. On X, analysts are connecting the new figure to OpenAI's massive fundraising — like the $122 billion round in March — and questioning the path to profitability. The narrative is that even the market leader is struggling to make the unit economics work, despite massive enterprise adoption.
Some observers see this as a simple accounting clarification, noting the difference in how OpenAI and Anthropic calculate revenue. The more bearish take, however, points to the delayed IPO, now slated for early 2027. This correction, combined with the IPO push, has some investors wondering if the growth story is less straightforward than it appeared.
Sailfish's take
We're not surprised by the correction. The economics of frontier models have always been brutal, and the public numbers were starting to feel inflated. For builders, this isn't just trivia — it's a stability warning. When a foundational provider's revenue story shifts this much, you have to plan for turbulence.
This means price hikes, API changes, or sudden deprecations are all more likely as they chase profitability. We've seen this playbook before. If your entire stack depends on a single provider, you've built on a fault line. This is the week to start properly testing a fallback model from another lab. Don't wait for the first outage or price change to diversify.