Meta’s AI Lab Delivers First Models After Six Months; Stock Falls 8.5% YTD
Meta’s Superintelligence Labs has delivered its first in-house AI models just six months after the lab’s launch, according to CTO Andrew Bosworth at Davos. Meanwhile, Meta stock is down 8.5% year-to-date and 20% since October, despite 26% revenue growth driven by AI-powered ads generating a $60B annual run rate.
1. Meta’s Superintelligence Labs Rolls Out First In-House AI Models
Meta’s newly formed Superintelligence Labs team has delivered its first batch of internally developed AI models just six months after its inception. CTO Andrew Bosworth revealed at the World Economic Forum in Davos that these early models, codenamed projects Avocado and Mango, have shown “very good” performance in preliminary tests. While Avocado targets advanced text-generation use cases and Mango focuses on image and video processing, both are undergoing extensive post-training refinement to ensure they can be integrated seamlessly into Meta’s suite of products. This milestone underscores Meta’s pivot from relying on external partners toward building proprietary AI infrastructure and IP at scale.
2. AI-Driven Ad Revenue Fuels 26% Growth in Q3
Meta reported 26% year-over-year revenue growth in its third quarter, up from 22% in the prior period, driven largely by enhancements in its AI-powered ads ranking systems. CEO Mark Zuckerberg highlighted that the annualized run rate for AI-powered ad tools has now surpassed $60 billion, marking a critical inflection point in monetizing its machine-learning investments. Meanwhile, the company’s newly launched AI assistant has exceeded 1 billion monthly active users, demonstrating strong engagement across Facebook, Instagram and WhatsApp. Operating income rose 18% to $20.5 billion, and free cash flow reached nearly $11 billion in the quarter, even as capital expenditures are slated to rise to $70–72 billion in 2025 and are expected to top $100 billion in 2026 to support its AI compute ambitions.
3. European Regulations Pose Hurdles for AI Rollouts
At the Davos meeting, Nicola Mendelsohn, head of Meta’s global business group, warned that the current regulatory landscape in Europe is making it significantly harder to deploy new AI features and products. She noted that compliance burdens and fragmented data-use rules are slowing down product launches on the continent, causing frustration among both Meta’s engineering teams and its business customers. Mendelsohn emphasized that without clearer, unified guidelines, Europe risks falling behind competitors in AI innovation and adoption, potentially constraining Meta’s ability to scale emerging offerings such as generative-AI tools and AR/VR experiences in the region.