Network effects are the engine of social platform success. Without them, no amount of investment can sustain a social product. Meta has shut down Horizon Worlds on VR — off the Quest store by March, terminated on June 15 — after close to $80 billion in losses. Mark Zuckerberg’s metaverse never generated the network effects that would have made it self-sustaining and self-growing. Close to $80 billion confirmed that network effects cannot be purchased — they must be earned.
Network effects occur when a platform’s value increases with each additional user, creating a self-reinforcing growth cycle. Once a social platform achieves critical mass, each new user both benefits from existing users and makes the platform more valuable for them. The cycle, once established, drives exponential growth that can transform a niche product into a global platform without proportionate increases in marketing or infrastructure investment.
Horizon Worlds never established the cycle. Its few hundred thousand monthly users created social spaces that were genuine communities but not the dense, vibrant social environments that generate network effect-driven growth. When a new user entered the platform and found virtual spaces with limited population, the experience did not generate the value that motivates return visits and word-of-mouth adoption. The network effect engine did not start.
Reality Labs spent close to $80 billion attempting to provide the equivalent of network effects through investment — creating content, events, and features that would make the platform valuable enough for each user even before critical mass was reached. The approach is sound in principle; it failed in practice because the substitute for network effects — manufactured value — could not replicate the organic value that genuine network effects produce.
Layoffs of more than 1,000 Reality Labs employees in early 2025 and the formal AI pivot acknowledged the failure. AI products are generating network effects differently but genuinely — the more people use AI tools, the more the models improve, and the more valuable they become for all users. This virtuous cycle, rooted in the accumulation of training data and usage patterns, provides AI with something closer to genuine network effects than the metaverse ever achieved.