Meta plans AI agent network Muse with $135 billion infrastructure spend

Meta’s AI agent network ambitions

Mark Zuckerberg, CEO of Meta, has detailed plans to build a large-scale network of AI agents under the name Muse. The goal is to serve billions of users globally, with the infrastructure required to scale this network already driving significant investments. Meta’s spending on AI infrastructure is projected to reach at least $135 billion this year, reflecting the company’s commitment to expanding its AI capabilities.

Privacy and legal challenges

Muse, which debuted in early September, has already faced scrutiny over privacy issues. Reports allege unauthorized analysis of private user texts and leaks of sensitive information, including a user’s home address. These incidents add to Meta’s existing legal troubles: a New Mexico judge recently ordered the company to pay $25 billion for misleading consumers about data privacy in the Cambridge Analytica scandal. Meta plans to appeal the verdict.

Separately, Meta addressed a major security vulnerability that allowed hackers to hijack user accounts, though details about the patch remain limited.

Financial and operational pressures

Meta’s aggressive AI investment strategy is straining its finances. The company reported a free cash flow deficit of $2.12 billion in the third quarter, a significant shift from previous quarters. This financial pressure coincides with operational challenges, including Amazon’s decision to block Muse agents from shopping on its platform. Amazon cited concerns over how Muse agents handled customer credentials and account data.

Stock restructuring and leadership commitments

To fund its AI ambitions without diluting Mark Zuckerberg’s control, Meta unveiled a new class of non-voting capital stock. The company will distribute this stock to existing shareholders in a two-for-one split next year. This move is partly aimed at mitigating potential financial risks, including a proposed 5% wealth tax on billionaires in California. Zuckerberg has reportedly threatened to relocate Meta’s headquarters if the tax passes in November’s election.

In a related development, Zuckerberg signed a Founder Alignment Agreement, committing to remain in California for the next eight years. The agreement ensures his continued leadership while allowing him to liquidate assets without losing control of Meta.

User growth and competition

Meta’s AI and social media initiatives are showing mixed results in user growth. Over 10 million people in the U.S. and Canada have set up personal Muse agents since its launch. Meanwhile, Meta’s Threads platform surpassed X (formerly Twitter) in monthly active users during the past quarter, signaling a shift in the competitive landscape for text-based social media.

AI infrastructure and scalability

Zuckerberg emphasized that Muse is "very compute-intensive" to run, and scaling it to a billion users will require unprecedented investment. Meta’s $135 billion budget for AI infrastructure this year underscores the company’s belief in the long-term potential of AI agents, despite the technical and financial challenges involved.

What this means

Lazyfounder analysis — our interpretation, not reported fact.

Meta’s push to build a network of AI agents at global scale is a high-stakes gamble that could redefine how startups and operators think about AI adoption. For founders, the story is a case study in the trade-offs between rapid scaling and the risks of privacy breaches, regulatory backlash, and financial strain. The $135 billion spend on AI infrastructure alone signals that Meta is betting big on AI agents becoming a core part of consumer and business interactions—but the legal and operational challenges it faces may serve as cautionary tales for others.

For operators, Meta’s struggles with Amazon’s ban on Muse agents highlight the risks of integrating AI into third-party platforms. The decision underscores how dependent AI agents are on partnerships, and how quickly those partnerships can sour over security or data-handling concerns. Startups working on AI-driven e-commerce or customer service tools should note the importance of robust credential management and transparent data practices.

The stock restructuring and financial pressures also offer lessons for founders navigating investor relations and corporate governance. Meta’s creation of a non-voting stock class to fund AI ambitions without diluting control is a creative but controversial approach. It reflects the tension between maintaining founder vision and meeting fiduciary responsibilities, a balance many startups will face as they scale.

Finally, the privacy accusations and the $25 billion penalty serve as a stark reminder of the regulatory minefield around AI and data. Startups building AI agents or products must prioritize compliance and user trust from day one, as retroactive fixes can be costly—both financially and reputationally.

Related on Lazyfounder

• Meta
• Amazon
• More ai news
• Arm’s AGI CPU Outperforms AMD in Agentic AI Benchmarks, Signal65 Reports