Meta on Pace to Surpass Google in AI Within Six Months, SemiAnalysis Reports

Meta Platforms (META) is positioned to leapfrog Google in the development of frontier artificial intelligence models within the next six months, according to a new report from boutique research firm SemiAnalysis. The analysis attributes this potential shift to Meta's aggressive capital deployment and strategic restructuring focused on proprietary data, talent acquisition, and a massive expansion of its computing infrastructure.

A Three-Pronged Strategy

SemiAnalysis highlights Meta's efforts to build a sustainable advantage through a highly sophisticated, proprietary data pipeline. The company has reportedly reallocated 3,000 engineers to create an in-house reinforcement learning (RL) environment by tracking internal employee workflows. This strategy is designed to generate training data that is unavailable to competitors relying on public or commercial sources.

This data initiative is complemented by a significant talent acquisition campaign. The report notes that Meta has spent billions to hire top-tier researchers from competitors like OpenAI, Anthropic, and Scale AI, including a reported $14.3 billion investment related to Scale AI, to assemble a team with expertise in advanced AI development.

On the product front, Meta recently released developer access to its Muse Spark 1.1 model. The move positions it as a direct competitor to the paid API models from rivals like Anthropic and OpenAI, underscoring its ambition to deliver powerful, multi-step agentic AI.

Unprecedented Infrastructure Buildout

Underpinning Meta's strategy is an unprecedented investment in physical infrastructure. SemiAnalysis projects Meta will surpass both OpenAI and Anthropic in total AI compute capacity by the end of the year. The company is simultaneously constructing five gigawatt-scale "titan" datacenter clusters connected by a custom networking architecture.

Ad

A recent Reuters report, citing an internal memo, underscored the scale of this hardware expansion:

• Meta plans to spend up to $145 billion on AI infrastructure this year.
• The company aims to deploy 7 gigawatts of computing power in 2026.
• This capacity is slated to double to 14 gigawatts in 2027.

To support this growth and reduce long-term costs, Meta is set to begin production of its custom AI chip, codenamed "Iris," in September, according to the Reuters report. The chip was reportedly designed with Broadcom and will be manufactured by TSMC.

Market Impact and Outlook

Wall Street reacted to the developments, with shares of Meta Platforms Inc. (NASDAQ:META) climbing 4.7% in recent trading. Conversely, shares of Alphabet Inc. (NASDAQ:GOOGL) fell by approximately 1%, reflecting investor concern over Google's competitive positioning.

While the initial performance benchmarks for Muse Spark have lagged some open-source peers, SemiAnalysis argues this view is shortsighted. The firm states that for Meta's AI division, "what matters... is the slope, not the intercept," emphasizing that the rapid pace of improvement is more critical than its current standing. The report concludes that if Meta maintains its current investment trajectory, Google risks being permanently displaced from the top tier of global AI hyperscalers.