Meta’s Superintelligence Labs Delivers First AI Models After Six Months
Meta’s Superintelligence Labs has completed initial development of in-house AI models roughly six months after its launch, according to CTO Andrew Bosworth. Early internally tested models, reportedly codenamed Avocado and Mango, show strong foundational performance but still need extensive post-training refinement before rollout across Meta’s services.
1. UK Regulator Opens Formal Inquiry Into WhatsApp Data Requests
On Friday, Britain’s communications regulator, Ofcom, announced it has launched an investigation into Meta Platforms’ compliance with information requests related to WhatsApp for one of its ongoing market reviews. The probe will examine whether Meta provided full and accurate data in response to official requests concerning messaging volume, content moderation practices and user metadata. Ofcom cited the importance of these disclosures in assessing competition and consumer protection across messaging services used by over 40 million UK residents. Investors should monitor potential enforcement actions or fines that could impact Meta’s regulatory expenses and reputation in one of its largest advertising markets.
2. Valuation Pullback Presents Attractive Entry Point
After a recent share pullback that has left Meta trading at roughly 21 times projected 2025 earnings and 11 times enterprise-value to EBITDA—well below its five-year averages—the company appears attractively valued. Consensus forecasts call for fourth-quarter revenue growth in the mid-20s percentage range and EPS above $8.50, which would exceed street estimates. While near-term margins may face headwinds from stepped-up AI computing investments and rising depreciation, Meta’s core advertising cash flow has generated free cash flow of nearly $11 billion in Q3 alone. Long-term drivers such as further AI ad-ranking improvements, ongoing metaverse development and increasing wearables adoption underpin a potential rerating back toward historical multiples.
3. Superintelligence Lab Delivers First Internal AI Models
Meta’s newly formed Superintelligence Labs team, less than six months into its mandate, has completed internal development of its first text and multimodal AI models. Chief Technology Officer Andrew Bosworth reported that these prototypes demonstrate promising capabilities in natural language understanding and image generation, validating the company’s pivot to in-house AI R&D. The models will undergo extensive post-training refinement to optimize performance and safety before broader roll-out to engineering and product teams. Successful internal deployment could accelerate improvements in ad targeting, recommendation engines and emerging offerings such as AI assistants across Meta’s family of apps, signaling a potential competitive edge over third-party model providers.