Meta's $2B Bet: Why AI's Next War is About Who Can Make the Robots Work

Meta just paid $2 billion for an AI startup that doesn't build its own models — but makes them work. The acquisition of Manus, a Singapore-based execution-layer startup, marks a strategic pivot toward orchestration-centric AI development.

Unlike traditional chatbot systems that prioritize conversational interfaces, Manus focuses on autonomous multi-step tasks like research, coding, and analysis. With 2 million waitlisted users and $100M ARR in 8 months, the startup processes 147 trillion tokens and creates 80 million virtual computers to power production-level workflows.

Manus’s approach integrates third-party AI models from Anthropic, Alibaba, and others instead of training proprietary models. This execution-first strategy enables complex use cases such as long-form research reports, data-driven visualizations, and multi-country travel planning.

For example, a user can request a market research report comparing MacBook models across 10 countries, and Manus will autonomously gather pricing data, analyze technical specs, and generate a comparative visualization—all in under 10 minutes. Traditional chatbots would require manual intervention for each step.

Meta plans to embed Manus into enterprise tools like Meta Business Suite, targeting small businesses rather than standalone software.

The integration will leverage Manus’s ability to reduce task completion time by 4x in version 1.5, as seen in travel planning scenarios where users previously spent hours coordinating flights, visas, and local guides.

Now, Manus automates these steps using a combination of external APIs and AI models. However, adoption remains skewed toward enterprise customers, with only 12% of waitlisted users currently active under a subscription model.