Meta is betting its future on personal AI agents, raising its 2026 capital expenditure forecast to as high as $145 billion to fund the initiative.
Meta Platforms is preparing to enter the agentic AI race with a consumer-focused assistant, backed by a capital expenditure plan now forecast between $125 billion and $145 billion for 2026. The move, reported by the Financial Times on May 5, follows the launch of its Muse Spark large language model and aims to build an AI that is more accessible than offerings from competitors like OpenAI and Google.
The goal is to deliver agents that can understand a person’s goals and “work day and night” to help achieve them, CEO Mark Zuckerberg said on the company's recent earnings call, positioning the technology as a tool to amplify people rather than replace them.
The strategy is already showing early signs of traction in the business world, where weekly conversations with Meta’s business AIs have surged 10x since the beginning of the year to over 10 million. This comes as the company posted strong Q1 financials, with revenue climbing 33% year-over-year to $56.3 billion, fueled by AI-driven recommendation improvements that boosted Reels time spent on Instagram by 10%.
The enormous investment required, however, presents a significant risk. Meta recorded a $107 billion increase in multi-year contractual commitments for cloud and infrastructure in the first quarter alone. The scale of the spending has prompted caution from analysts, who question if the returns on AI can justify the cost, creating a high-stakes scenario for the social media giant.
The $145 Billion Question: Agentic AI
Meta’s increased capital expenditure, a jump from a previous estimate of $115-$135 billion, is primarily for servers, data centers, and custom silicon needed to train and deploy a global fleet of AI agents. While other major labs have focused on enterprise customers, Mizuho analysts noted Meta’s strategy is centered on consumer applications, a potentially vast but unproven market.
The company is partnering with Google, Perplexity, and OpenAI to shape the direction of "agentic commerce." However, some retail executives are skeptical about the universal applicability of AI agents. Wayfair CEO Niraj Shah argued that while agents may succeed with replenishable goods or commodities, they are less likely to dominate in categories like home furnishings where emotional and discovery-based shopping is key—a direct contrast to Meta's "be everywhere" ambition.
Engagement Gains Fuel AI Flywheel
The massive AI spend is predicated on Meta’s ability to translate investment into user engagement, and early data shows progress. Beyond the 10% lift in Reels consumption, total video watch time on Facebook grew by more than 8% in Q1, the largest quarterly gain in four years. Management attributes this to more advanced AI ranking models that have more than doubled the amount of same-day posts recommended to users.
This "AI flywheel"—where better models drive engagement, which generates data to improve the models—is critical to justifying the infrastructure outlay. The company is also deploying AI into its core ad systems, with improvements to its Lattice and GEM model architecture driving a 6% increase in conversion rates for certain ad types in the first quarter.
For investors, the strategy presents a clear trade-off. Mizuho recently lowered its price target on Meta from $850 to $835, citing the escalating capex and the need for a clearer product roadmap before the stock’s valuation multiple can expand. While maintaining an Outperform rating, the firm warned that if growth peaks without definitive progress, the stock could face pressure. Meta's path forward hinges on proving its $145 billion bet can create a new paradigm of consumer AI, and not just a new tier of expenses.
This article is for informational purposes only and does not constitute investment advice.
