Investing.com — Following a year of radical restructuring and aggressive capital deployment, Meta Superintelligence (MSL) is positioned to leapfrog Google in the frontier AI hierarchy within the next six months. According to a new report from boutique research firm SemiAnalysis, Mark Zuckerberg’s relentless pursuit of proprietary data, elite talent, and unprecedented compute infrastructure has effectively transformed the ecosystem into a race where Google has “faded dramatically”.
While competitors scramble for diminishing pools of public data, Meta has turned inward by tracking employee workflows and reallocating 3,000 engineers to build a massive, in-house reinforcement learning (RL) environment factory. SemiAnalysis notes that this internal supply chain gives MSL a highly sophisticated, proprietary data pipeline for training next-generation agents that commercial data brokers simply cannot replicate.
Further intensifying this push, Meta on Thursday released long-awaited developer access to its upgraded Muse Spark 1.1 model, pitting it directly against the paid API models of Anthropic and OpenAI. Touted as its most capable model for real-world coding and agentic tasks, Meta is pitching the upgrade as part of a broader mission to deliver “personal superintelligence” capable of executing multi-step tasks with minimal human intervention.
On the infrastructure front, SemiAnalysis projects Meta will surpass both OpenAI and Anthropic in total AI compute by year-end through the simultaneous construction of five gigawatt-scale “titan” datacenter clusters. Supported by a custom “AI-Backbone” networking architecture, this unprecedented hardware footprint allows Meta to scale complex training workloads asynchronously across locations separated by thousands of kilometers.
This aggressive hardware timeline was underscored by a Reuters report revealing that Meta plans to spend up to $145 billion on AI infrastructure this year, part of a massive global buildout. According to Reuters, citing an internal memo, the company plans to deploy 7 gigawatts of computing power in 2026 and double that capacity to 14 gigawatts in 2027.
To support this expansion and lower computing costs, Meta will begin production of its custom AI chip, code-named “Iris,” in September, according to the Reuters report. Designed alongside Broadcom and manufactured by TSMC, the silicon cleared bug testing in just six weeks and is backed by multi-year supply agreements with Samsung, SanDisk, and Sumitomo Electric.
