SemiAnalysis on the State of AI: A Two-Horse Race, Google's Slide to Fifth, and the $60B Revenue Inversion

The AI industry has undergone a dramatic power shift in 2026, according to the latest research from semiconductor and AI analysis firm SemiAnalysis. Their July 2026 reports paint a picture of a sector that has consolidated into a duopoly, seen an incumbent crumble, and discovered its killer use case in software development.

The OpenAI–Anthropic Duopoly: A Two-Horse Race

SemiAnalysis states that the AI industry has consolidated into a two-horse race between OpenAI and Anthropic . OpenAI mounted a strong comeback with its new model (reportedly version 5.6) and the Codex coding tool, while Anthropic has surged on enterprise adoption and the viral growth of Claude Code . Together, these two labs now dominate the frontier of AI capabilities and commercial revenue .

Google's Decline to Fifth Place: A Failure of Compute Strategy

Perhaps the most striking finding from SemiAnalysis is Google's precipitous fall. The firm assesses that Google has dropped to fifth place among AI labs globally . The root cause is described as "compute binding and strategic errors" — specifically, a failed approach to TPU leasing that locked Google into long-term rental agreements without adequate "clawback" provisions, limiting its own access to compute .

This strategic misstep triggered a talent exodus. High-profile researchers have departed in a wave: Transformer co-author Noam Shazeer, Nobel laureate John Jumper, and Gemini researchers Jonas Adler and Alexander Pritzel have all left for rivals . Alphabet lost up to $270 billion in market cap in a single day following these departures . Google dropped out of the top 5 on the Artificial Analysis Intelligence Index for the first time in July 2026 .

Programming: The Dominant Revenue Driver Consuming 70% of Tokens

SemiAnalysis's latest interview reveals that programming use cases now consume 70% of all tokens across major AI platforms . This dominance is forcing tech giants to end subscription subsidies and pivot toward high-margin API monetization, as labs push usage toward API channels with gross margins of 85%+ . Claude Code alone has reached $2.5 billion in annualized revenue just nine months after launch .

Financial Metrics: The Great Inversion

SemiAnalysis's Tokenomics Model provides detailed financial portraits of both leading labs, and the numbers show a dramatic reversal.

Anthropic has experienced explosive growth:

• ARR surged from approximately $9 billion at the end of 2025 to over $60 billion by July 2026 .
• Projected Q3 2026 GAAP EBIT profit exceeding $1 billion (operating margin ~6%) .
• Inference gross margins exceeding 70% .
• Q2 2026 revenue projected at $10.9 billion, up 127% quarter-over-quarter from Q1's $4.8 billion .

OpenAI, meanwhile, has seen its relative position weaken:

• Was pacing at roughly $25 billion annualized run-rate earlier in 2026, up from $13.1 billion in full-year 2025 .
• Monthly revenue reached approximately $2 billion ($24 billion annualized) by April 2026 .
• Still spending heavily on compute, with a Q2 2026 EBIT profit margin of approximately -100% .

SemiAnalysis notes the financial dynamics inverted in Q2 2026, with Anthropic surpassing OpenAI in revenue and posting its first profit while OpenAI continues to invest heavily in compute infrastructure .

Greg Brockman's Remarks: Competition, AGI, and the Code Revolution

OpenAI President Greg Brockman has offered several notable comments on the state of the industry. In a July 23, 2026 Bloomberg interview, he said it is still "the early innings of where this industry goes" regarding competition with Anthropic . Earlier in June 2026, he stated that "what's more important than AGI is actual capability" and that the most important challenge is to "connect AI models with the world" .

At Sequoia's AI Ascent 2026 conference, Brockman suggested OpenAI is about 80% of the way to AGI . He also revealed that AI now writes approximately 80% of OpenAI's own code — up from 20% just six months prior .

Meta Overtaking Google?

The available SemiAnalysis sources do not directly address the possibility of Meta overtaking Google. While Google's drop to fifth place leaves an opening for well-funded competitors, there is insufficient evidence in the current SemiAnalysis materials to confirm or deny a specific Meta-overtaking-Google thesis . Additional direct sourcing from SemiAnalysis's own research on Meta's AI positioning would be required to make that determination.

---

This article is based on SemiAnalysis reports and interviews from July 2026, including their latest interview published on July 23, 2026, and their Tokenomics Model report from July 8, 2026. All financial figures are sourced from SemiAnalysis's proprietary models and public reporting.