Meta’s newly formed Superintelligence Labs team, launched six months ago, has completed its first set of in-house AI models, according to CTO Andrew Bosworth. Presented at the World Economic Forum in Davos, these early-generation models—codenamed Avocado (text) and Mango (vision)—demonstrated “very good” performance on internal benchmarks. Bosworth noted that post-training integration and optimization will extend over the coming quarters to enable deployment across Meta’s suite of social apps and eventual consumer release. This milestone underscores Meta’s shift from third-party reliance to self-sufficiency in foundational AI technology.

After rising 20% through October, Meta’s shares have declined 8.5% year-to-date, reflecting investor concerns over massive capital expenditures. In 2025 Meta plans to spend $70–72 billion on data-center infrastructure and expects to exceed $100 billion in 2026, driven primarily by AI compute needs. Yet the company reported third-quarter revenue growth of 26% year-over-year (up from 22% in Q2), powered by a $60 billion annualized run rate from AI-driven ad ranking and over 1 billion monthly active users for its Meta AI assistant. Operating income rose 18% to $20.5 billion in Q3, with free cash flow of nearly $11 billion, supporting a current valuation of 27 times trailing earnings (21 times forward). Management forecasts fourth-quarter revenue of $56–59 billion, underpinning the argument that the current share price dip represents a long-term buying opportunity.

Nicola Mendelsohn, head of Meta’s global business group, told Bloomberg at Davos that Europe’s stringent AI regulations have slowed product launches and frustrated corporate customers. Mendelsohn cited overlapping compliance requirements across data protection, content moderation and algorithmic transparency as key pain points, requiring multiple rounds of legal and technical reviews. She warned that delays of six to nine months on some AI features risk eroding Meta’s competitive edge in global product rollouts and urged harmonization of rules to accelerate innovation.

Meta recently cut 1,500 positions—about 10% of its 15,000-employee Reality Labs division—in its latest cost-reduction drive. The layoffs, concentrated in engineering and product teams, follow underwhelming revenue from hardware headsets and developer engagement below initial forecasts. A separate Washington State filing revealed 331 staffers affected at two Puget Sound offices. The move underscores Meta’s strategic pivot away from standalone metaverse hardware toward prioritizing AI-driven social experiences and more immediate revenue streams.