SemiAnalysis: Meta AI is expected to surpass Google within half a year, forming a three-way rivalry with OpenAI and Anthropic.

The global landscape of cutting-edge AI competition is undergoing a profound reshaping.

SemiAnalysis's latest report points out that, after a year of aggressive capital investment and organizational restructuring, Meta's Meta Superintelligence (MSL) is expected to surpass Google in cutting-edge AI capabilities within the next six months.

Meanwhile, Anthropic, thanks to the explosive growth of Claude Code in the enterprise market, has established a dual advantage in profitability and growth rate, and secretly submitted its IPO application in June. The power structure of the AI industry is evolving from a Google vs. OpenAI rivalry to a new order where Meta, OpenAI, and Anthropic form a three-way stalemate.

On the market front, these dynamics have triggered immediate responses. Meta’s stock price rose about 4% in a day, while Alphabet’s stock price fell about 1%, reflecting the market’s growing concerns over Google's position in the AI race. Meanwhile, Musk publicly stated on social platform X that Anthropic is "clearly" the current leader in AI, adding another variable to the discussion on the competitive landscape.

SemiAnalysis's earlier in-depth financial analysis of Anthropic's IPO reveals that, the company is expected to achieve GAAP EBIT profits exceeding $1 billion in Q3 2026, while OpenAI's EBIT profit margin remains at -100%. Accordingly, SemiAnalysis gives Anthropic a benchmark valuation of $6 trillion, and believes Anthropic should go public ahead of OpenAI to further consolidate its leading position in frontier model fields through capital advantage.

Meta: Advancing on Three Fronts—Compute, Data, Talent

The core judgment of the SemiAnalysis report is: Meta’s expansion in compute scale will allow it to exceed the combined AI compute capacity of OpenAI and Anthropic by the end of the year.

According to Reuters citing an internal memo, Meta plans to invest up to $145 billion in AI infrastructure this year, deploy 7 GW of compute by 2026, and double that number to 14 GW by 2027. Supporting this goal are five gigawatt-level “titan” hyperscale data center clusters being constructed simultaneously, as well as Meta’s in-house “AI-Backbone” network architecture—which allows Meta to asynchronously scale complex training tasks over thousands of kilometers of geographic distance.

On the chip front, according to Reuters, Meta will begin mass production in September of its self-developed AI chip, codenamed "Iris", jointly designed with Broadcom and manufactured by TSMC. Vulnerability testing was completed in just six weeks, and Meta has signed multi-year supply agreements with Samsung, SanDisk, and Sumitomo Electric.

In terms of data and talent, Meta has reassigned 3,000 engineers to internal reinforcement learning (RL) environment factories, creating proprietary data pipelines that commercial data providers cannot replicate. Meanwhile, Meta invested $14.3 billion in Scale AI, facilitating large-scale poaching of top researchers from OpenAI, Anthropic, and others.

SemiAnalysis believes that, evaluating MSL’s current benchmark performance is "seeing the trees but not the forest"; what truly matters is "the slope rather than the intercept"—if Zuckerberg maintains the current level of capital investment, Google could be permanently excluded from the first tier of global AI hyperscale players.

Anthropic: B2B Profitability Model Establishes Leadership

If Meta's strengths lie in compute and infrastructure, Anthropic's moat is built on business model and financial quality.

SemiAnalysis data shows Claude Code currently accounts for over 7% of all code commits on GitHub. This product has driven Anthropic’s ARR from $9 billion at the end of 2025 to $30 billion in a single quarter in Q1 2026—$3 billion added in January, $7 billion in February, $11 billion in March. So far, Anthropic’s ARR has exceeded $60 billion, with net new ARR of over $10 billion per month.

Anthropic’s financial structure contrasts sharply with OpenAI. About 75-85% of Anthropic's revenue comes from usage-based API business, while OpenAI still has over 65% of its revenue from subscription models in Q1 2026, with consumer subscriptions accounting for about 40%. SemiAnalysis estimates OpenAI needs to serve over 900 million free users, with a monthly per-user cost of roughly $0.70, dragging its overall gross margin down by 20-30%.

In terms of profitability, SemiAnalysis expects Anthropic to achieve GAAP EBIT of over $1 billion (with a margin of about 6%) in Q3 2026, while OpenAI's EBIT margin remains at -100%. Anthropic CFO Krishna Rao previously disclosed that the company's net dollar retention (NDR) is as high as 500%, meaning old customers who contributed $2 billion in ARR a year ago now contribute $12 billion ARR. Based on this, SemiAnalysis predicts that if Anthropic can accelerate monthly net new ARR to $15 billion, its ARR by the end of 2027 is expected to reach $300 billion, corresponding to an enterprise value of $6 trillion.

Google: From Leader to One Being Chased

In this three-party contest, Google’s situation is the most delicate.

SemiAnalysis bluntly states in its report that, Google has "significantly regressed" in frontier AI competition, and predicts Meta will surpass Google within six months. Musk named Anthropic as the "clear" leader in current AI, not including Google in the top tier discussion.

SemiAnalysis's analytical framework shows that the decisive variables in frontier AI competition have shifted from model capability alone to a comprehensive contest of compute scale, business model, and capital acquisition capability.

In these three dimensions, Meta is quickly catching up, Anthropic has established a lead, and Google faces pressure from both directions.

Risk Warning and Disclaimer ClauseThe market carries risks; investments should be made cautiously. This article does not constitute personal investment advice, nor does it take into account individual users' specific investment goals, financial situations, or needs. Users should consider whether any opinions, views, or conclusions in this article are suitable for their particular circumstances. Investing based on this information is at your own risk.