Meta's Superintelligence Lab One Year Update: From Aggressive Hiring to Surpassing OpenAI in Compute

Meta's strategic layout in AI can be traced back to 2024, when the company decided to establish an independent superintelligence lab and began an unprecedented talent war. SemiAnalysis's report details Meta's recruitment strategy: 'Meta became famous for offering AI researchers compensation packages that would make Patrick Mahomes jealous.' This cost-no-object recruitment approach is extremely rare in tech history, reflecting Meta's extreme emphasis on superintelligence R&D. The report notes that MSL's recruitment targets extend beyond OpenAI, broadly covering core researchers from top AI labs including Anthropic and Google DeepMind.

In terms of specific talent acquisition, the report lists a series of heavyweight researchers who have joined. Shengjia Zhao is a well-known scholar in machine learning who previously enjoyed great reputation in academia; Trapit Bansal has deep expertise in computer vision and multimodal AI; Joel Pobar is a senior engineering lead at OpenAI who participated in the development of multiple core projects; Jack Rae is an important researcher at DeepMind with outstanding contributions in reinforcement learning. The addition of these talents has enabled MSL to establish complete capabilities covering fundamental research, engineering implementation, and productization in just one year.

Compute is another critical dimension of AI competition. SemiAnalysis's report analyzes Meta's compute advantage from a 'Tokenomics' perspective. Unlike OpenAI and Anthropic, Meta has a balance sheet befitting a hyperscaler. More importantly, Meta doesn't have the business burden of needing to rent out large amounts of compute to cloud customers like Google does. Combined with CEO Zuckerberg's willingness to accept negative free cash flow as a strategic choice, Meta has the ability to bring up more internal AI compute than anyone else in the world. SemiAnalysis's new Tokenomics model projects that Meta will have more AI compute than OpenAI and Anthropic combined by the end of this year.

On the product front, Meta's superintelligence lab has already begun producing results. In April 2026, Meta released the Muse Spark model, breaking its silence in the AI field. Subsequently, the company launched a series of products: the Muse Image generation model (though controversial for opting in public Instagram accounts by default), the Muse Spark 1.1 iteration, and Meta's first paid AI model. When launching the paid model, CEO Zuckerberg called other AI labs' pricing 'very extreme' and promised Meta would enter the market at 'very low prices.' This aggressive pricing strategy is consistent with Meta's 'free + advertising' business model in social media, showing its attempt to gain market share through scale effects rather than high profit margins.

From a competitive landscape perspective, Meta's entry is profoundly changing the balance of power in the AI industry. SemiAnalysis's report notes that frontier AI increasingly feels like a 'two-horse race' between OpenAI and Anthropic. Google had a brief moment in the spotlight with Gemini 3 Pro and Nano Banana, but has since 'faded dramatically.' Despite their Windsurf acquisition, they're far from a compelling agentic coding product, and 3.5 Flash performs far worse than GPT 5.6 and Opus 4.8 in real-world scenarios. Microsoft has completely blown their early lead with GitHub Copilot and failed to effectively leverage their access to OpenAI IP. SpaceXAI is selling $26 billion a year worth of GPUs to Anthropic and Google, while Chinese labs are simply too compute-poor to truly reach the frontier.

Meta's strategic advantage also lies in its unique business model. Unlike OpenAI and Anthropic, which rely on subscription revenue, Meta's core business is social media advertising. This means AI models can be viewed as tools to enhance core businesses (Instagram, Facebook, WhatsApp) rather than standalone profit products. Instagram ad revenue can fund substantial compute — as SemiAnalysis's report points out, 'Instagram ads can fund a lot of compute.' This business model enables Meta to make long-term investments in AI without needing to prove commercial returns as urgently as OpenAI must.

For users following AI industry developments, Meta's rise means more choices and lower prices. As Meta enters the market with aggressive pricing, AI service prices across the industry are rapidly declining. Developers can use Evergreen Tools' API成本计算器 to compare prices across different AI providers and choose the optimal solution. Meanwhile, 图片压缩工具 and 图片尺寸调整工具 can help users better process AI-generated image content.