OpenAI Crosses $25B Annual Revenue, Eyes IPO — Anthropic Closes the Gap
OpenAI has hit $25B in annualized revenue — roughly $2B per month — and is now actively exploring a late-2026 IPO. Meanwhile, Anthropic is approaching $19B annualized, and secondary-market investors are quietly shifting capital from OpenAI shares to Anthropic positions.
The Numbers Behind the Surge
OpenAI's $25B run rate marks a staggering acceleration from under $4B just 18 months ago. The company is reportedly generating $2B in monthly revenue, driven by ChatGPT Plus subscriptions and enterprise API consumption. An IPO — potentially in late 2026 — would test whether public markets assign the same premium that private rounds have.
Anthropic's trajectory is equally striking. At roughly $19B annualized, the gap between the two leaders has narrowed considerably. Secondary-market demand for OpenAI shares has cooled, with multiple investors pivoting allocations toward Anthropic — a signal that the market no longer views this as a one-horse race.
Model Benchmarks: A Three-Way Race
Revenue competition is mirrored in model performance. GPT-5.4 (released March 5) set new computer-use benchmark records and posted a GDPval score of 83%. Claude Sonnet 4.6 holds the GDPval-AA Elo lead at 1,633 points. Google's Gemini 3.1 Pro pushed GPQA Diamond to 94.3%.
Model Benchmark Snapshot (as of April 2026)
─────────────────────────────────────────────
GPT-5.4 │ GDPval 83% │ Computer-use records
Claude Sonnet 4.6│ GDPval-AA Elo 1,633 │ Reasoning + agentic tasks
Gemini 3.1 Pro │ GPQA Diamond 94.3% │ Scientific QA
No single vendor dominates every benchmark. The practical implication: enterprises selecting foundation models must now evaluate across multiple axes — reasoning depth, agentic reliability, domain-specific accuracy — rather than defaulting to a single provider.
Startup Ecosystem: Seed Rounds Reflect the Hype
AI startup seed valuations continue to climb. $10M seed rounds at $40-45M post-money are becoming routine for teams with strong model fine-tuning or vertical-AI theses. The funding environment rewards speed and defensibility — companies that can demonstrate production-grade AI deployments, not just demos.
Industry Implications
-
Multi-model strategies are now baseline. With three credible frontier labs shipping competitive models on overlapping timelines, vendor lock-in carries real risk. Enterprises should architect for model-agnostic inference.
-
An OpenAI IPO reprices the entire sector. Public-market scrutiny will force transparency on margins, churn, and compute costs — metrics that have been opaque in private rounds. The resulting valuation anchor will cascade to every AI startup raising capital.
-
Voice AI benefits directly from model competition. Faster, cheaper, more accurate foundation models compress the cost of real-time voice pipelines — STT, LLM reasoning, TTS — making production-grade voice agents viable at price points that were impossible 12 months ago.
-
Secondary-market rotation signals a maturing investor thesis. Capital flowing from OpenAI to Anthropic suggests investors are diversifying bets across the frontier, not just chasing the incumbent. This benefits the broader ecosystem by validating alternative approaches to AI safety and capability.
