SemiAnalysis, a boutique research firm, argues that Meta Superintelligence (MSL) is on track to surpass Google among frontier AI contenders within the next six months. The report credits an intensive internal restructuring and a heavy concentration of capital at Meta with creating a competitive environment in which, the firm says, Google has "faded dramatically."
At the center of SemiAnalysis's assessment is Meta's pivot away from public data sources toward a proprietary, internal data supply chain. The company has reportedly reassigned roughly 3,000 engineers to construct a large-scale, in-house reinforcement learning (RL) environment factory. SemiAnalysis describes that internal capability as a sophisticated data pipeline for training next-generation agents, one that it believes commercial data brokers cannot replicate.
Adding to that capability stack, Meta opened developer access to an updated model, Muse Spark 1.1. The company markets Muse Spark 1.1 as a marked improvement for real-world coding and agentic tasks, positioning the upgrade as part of a broader effort to deliver "personal superintelligence" that can complete multi-step assignments with limited human oversight. The release sets Muse Spark 1.1 in direct competition with paid API models from Anthropic and OpenAI.
On the infrastructure front, SemiAnalysis projects that Meta will exceed the total AI compute of both OpenAI and Anthropic by the end of the year. That forecast rests on Meta building five gigawatt-scale "titan" datacenter clusters simultaneously, supported by a custom "AI-Backbone" networking design intended to let the company scale training workloads asynchronously across sites separated by thousands of kilometers.
The aggressive hardware timeline is backed, according to the report, by a Reuters account that Meta plans to spend as much as $145 billion on AI infrastructure this year as part of a global expansion. Citing an internal memo, Reuters reported plans to deploy 7 gigawatts of compute in 2026 and to double that figure to 14 gigawatts in 2027.
To drive down operating costs and secure capacity, Meta intends to begin production of a custom AI chip named "Iris" in September. The chip was developed in partnership with Broadcom and is to be manufactured by TSMC. SemiAnalysis relays that Iris completed bug testing in six weeks and that Meta has established multi-year supply agreements with Samsung, SanDisk, and Sumitomo Electric.
The company has paired this buildout with an expansive talent recruitment effort. SemiAnalysis documents billions in recruiting spend, including a $14.3 billion Scale AI investment intended to attract top researchers from OpenAI, Anthropic, and Scale AI itself. That campaign, SemiAnalysis contends, has assembled an "AI superteam" capable of turning Meta's large-scale compute into frontier model performance.
Wall Street's reaction was immediate. Meta Platforms Inc (NASDAQ:META) shares rose about 4% after recovering from an earlier intraday decline, while Alphabet Inc Class A (NASDAQ:GOOGL) shares fell roughly 1% during the session. The report frames these moves as market responses to the shifting balance of infrastructure and product milestones.
SemiAnalysis also addresses the comparative performance of Muse Spark. While early benchmark results placed Muse Spark below some open-source alternatives, the firm warns that isolated model scores can be misleading. "Missing the forest for the trees," SemiAnalysis writes, and argues the relevant metric for MSL is long-term trajectory: "what matters for MSL is the slope, not the intercept." The research house cautions that if Meta sustains its committed spending and execution, Google could be at risk of falling out of the top tier of global AI hyperscalers.
Context and implications
This analysis links together four principal elements supporting SemiAnalysis's forecast: a proprietary internal data pipeline built from employee workflow signals, the redeployment of 3,000 engineers to an RL environment factory, the release of Muse Spark 1.1 for developer access, and a planned multi-gigawatt hardware expansion supported by custom silicon and vendor agreements. Those elements, taken together, form the basis for the firm's view that Meta can scale frontier AI capabilities rapidly.
Market reaction
Traders responded to the combination of product and infrastructure news with divergent moves in equity markets, rewarding Meta's stock and applying pressure to Alphabet's Class A shares. SemiAnalysis interprets the shift as evidence that investors are recalibrating the relative positions of major AI players.
Limitations of the assessment
The report's conclusions rest on Meta's continued deployment of capital, talent and compute as described. SemiAnalysis emphasizes the importance of execution over time and frames its view around projected build and recruitment plans rather than a single benchmark score.