Anthropic Surpasses OpenAI in Revenue, Confidentially Files for IPO
Anthropic's enterprise focus drives significant revenue growth and IPO preparations, while Chinese AI models gain substantial traction in the US market, and new models are released across labs.
The story
The AI software landscape saw a notable shift this week, as Anthropic reportedly surpassed OpenAI in annualized revenue, crossing the $30 billion mark. This financial milestone underscores Anthropic's successful strategy in capturing the enterprise market, largely driven by its Claude Code offerings, and comes as the company confidentially filed for an Initial Public Offering on June 1st.
Meanwhile, a new report highlights a growing competitive dynamic from Chinese AI models, which now account for between 30% and 46% of US enterprise API token usage, according to CNBC. This suggests a pricing challenge for US frontier AI labs in the middle tier of enterprise workflows, pushing them to defend top-tier performance where cost is secondary. Elsewhere, labs continued to release new capabilities, including OpenAI's new fallback model and Microsoft's expanded in-house MAI family, signaling ongoing rapid development and diversification across the sector.
Who moved
Anthropic
What Changed: The company's annualized revenue run rate reportedly crossed $30 billion, surpassing OpenAI, and it confidentially filed for an IPO on June 1st.
Consequence: This signals a successful enterprise-focused strategy, particularly with its Claude Code offerings, and positions Anthropic for a public market debut.
OpenAI
What Changed: OpenAI rolled out GPT-5.5 Instant Mini in ChatGPT as a new fallback model, replacing GPT-5.3 Instant Mini.
Consequence: This update aims to improve user intent tracking, tone calibration, and reduce repetitive responses within the ChatGPT experience.
Microsoft
What Changed: Microsoft unveiled seven new in-house MAI (Microsoft AI) models at Build 2026, spanning image, speech, transcription, reasoning, and code.
Consequence: This move reduces Microsoft's dependency on partner OpenAI's technology and strengthens its independent model lineup for various applications.
SpaceXAI (formerly xAI)
What Changed: SpaceXAI is reportedly launching a new frontier AI model, developed jointly with the AI startup Cursor, which SpaceX is in the process of acquiring for $60 billion.
Consequence: This marks the first joint production with Cursor and is internally compared to models like Claude Opus 4.8 and OpenAI's GPT 5.5.
Bespoke Labs
What Changed: The AI post-training startup raised $40 million in a new funding round.
Consequence: The capital will support the development of infrastructure for reinforcement learning from human feedback (RLHF), preference data collection, and fine-tuning pipelines.
New models
GPT-5.5 Instant Mini
Lab: OpenAI
What: A new fallback model for ChatGPT, replacing GPT-5.3 Instant Mini.
Use: It is designed to better track user intent, calibrate tone, and avoid repetitive or overly structured responses for users who hit rate limits.
Claude Science Workbench
Lab: Anthropic
What: An extended version of the Claude Science application that connects Claude Opus 4.8 to over 60 scientific databases.
Use: Researchers can run multi-step scientific workflows, retrieving data, performing analysis, and synthesizing findings in a single session.
MAI-Thinking-1
Lab: Microsoft
What: Microsoft's new flagship reasoning model, scoring 97% on the AIME 25 test and 52.8% on SWE Bench Pro.
Use: It is designed for complex reasoning and code writing tasks, competitive with other advanced models.
MAI-Image-2.5
Lab: Microsoft
What: An image generation model that took second place on the Arena image editing leaderboard.
Use: It offers precise local edits, understanding of scene structure, lighting, and perspective, and preservation of facial features.
Unnamed frontier model
Lab: SpaceXAI / Cursor
What: A new frontier AI model developed jointly by SpaceXAI and Cursor.
Use: It is being compared internally to Anthropic's Claude Opus 4.8 and OpenAI's GPT 5.5, indicating advanced capabilities.
Market signals
Anthropic's annualized revenue run rate reportedly crossed $30 billion, surpassing OpenAI, and the company confidentially filed for an IPO on June 1st.
Implication: This validates Anthropic's enterprise-focused business model and signals strong investor confidence in its growth and profitability.
Global startups raised a record $510 billion in the first half of 2026, with over 70% of that capital flowing to AI companies, and OpenAI and Anthropic alone accounting for $217 billion (43% of the total).
Implication: This demonstrates unprecedented investor enthusiasm for AI, with a significant concentration of capital flowing to frontier model developers.
CNBC reported that Chinese AI models now account for 30-46% of US enterprise API token usage, with GLM-5.2 performing near Anthropic's Opus 4.8 at a fifth of the cost.
Implication: This indicates a significant penetration of cost-effective Chinese models into the US enterprise market, creating competitive pricing pressure for US labs.
As of July 8, 2026, Anthropic's Fable 5 model now requires usage credits at standard API rates, ending its inclusion in subscription plans.
Implication: This reflects Anthropic's strategy to monetize its advanced models and could increase costs for users who previously had bundled access.
Andrej Karpathy, a prominent AI researcher and educator, announced he has joined Anthropic.
Implication: This high-profile hire underscores Anthropic's ability to attract top talent and its continued investment in frontier large language model research and development.
Builders saying
Andrej Karpathy, AI Researcher. He announced joining Anthropic to work on the frontier of large language models and return to research and development, indicating a focus on core LLM advancements.
What we'll be watching
- The RAISE Summit in Paris continues through July 9, focusing on AI leadership and enterprise adoption.
- OpenAI is set to launch GPT-5.6, its most advanced AI model, on Thursday, July 9.
- The White House is expected to announce a voluntary framework for releasing frontier AI models as early as next week.
- AMD AI DevDay 2026 is scheduled, bringing together engineers and developers for technical sessions and workshops.
- C3.ai is estimated to announce its earnings between September 2, 2026, and September 8, 2026.
Reporting + analyst voices: grounded via Google Search at publish time.