ByteDance

$155 billion revenue in 2024 (38% growth) and $48 billion in Q2 2025—surpassing Meta's $46 billion—crowns ByteDance as the world's largest social media company by sales. But revenue metrics miss the biological insight: this is adaptive radiation from a single algorithm. The recommendation engine doesn't power 'an app'—it's the ancestral trait enabling speciation into TikTok (1.6 billion users), Douyin ($483 billion e-commerce GMV), Toutiao (news), and Doubao (60 million MAU AI chatbot). Same selective pressure (attention scarcity), different niches (short video, social commerce, news aggregation, conversational AI), one inherited advantage (hyper-personalized content delivery without friend graphs).

The $330-480 billion valuation (depending on private market trades) reflects network effects compounding through stigmergy: each user's behavior leaves digital pheromone trails that guide subsequent users' experiences. Unlike Facebook's social graph (who you know) or Google's knowledge graph (what you search), ByteDance's interest graph learns what you'll watch next based on billions of micro-signals—watch time, replays, completion rates, scroll velocity. This is swarm intelligence: no individual user understands the pattern, but aggregate behavior optimizes content distribution with eerie accuracy.

The AI infrastructure investment ($20B including $7B on Nvidia chips) targets competitive escalation: Doubao challenging DeepSeek requires the same recommendation algorithm adapted to conversational substrate. OmniHuman-1 (February 2025: realistic videos from single images) extends the pattern—generative content feeding the recommendation engine, creating closed-loop systems where AI creates, users consume, algorithms learn, content improves. This is phenotypic plasticity meeting cultural transmission: the platform adapts to user preferences while simultaneously shaping what users prefer.

At 150,000+ employees and ~$50 billion projected 2025 profit, ByteDance demonstrates how mycelial networks scale: distributed processing (each regional team adapting algorithm to local culture), resource sharing (core tech leveraged across products), and the ability to colonize new substrate (e-commerce, AI, gaming) faster than specialized competitors. TikTok Shop's $9B U.S. GMV (missing $17.5B target) shows limits—but also the r-selection approach: launch fast, iterate ruthlessly, scale what works, kill what doesn't.