Meta’s agentic AI assistant: a bet on billions of users and a US$145 billion bill

One specific use case is shopping. Meta plans to integrate a separate agentic shopping tool into Instagram, according to the same report . If that happens, Meta would be putting AI directly into one of its most commercially important social platforms, rather than treating it as a standalone assistant.

This is not yet a broad consumer launch. A Reuters report carried by The Star said the assistant is being tested internally by a group of staff, with the goal of developing a product similar to OpenClaw . Separately, The Information reported that Meta is training an internal AI agent codenamed Hatch, inspired by OpenClaw, with the aim of completing internal testing by the end of June . Reuters also reported that Meta did not immediately respond to a request for comment on the FT report .

Why “agentic AI” is more than a chatbot label

In this context, the important word is not “AI” but “do.” A typical chatbot mainly responds: it writes, summarizes, answers or suggests. The assistant Meta is reportedly building is framed around carrying out everyday tasks for users, not just producing text in response to prompts .

That makes the product challenge much bigger. If an AI assistant can act inside real app workflows — for example, through a shopping tool in Instagram — it starts to look less like a help box and more like a new interface layer for Meta’s apps . But more personalization and more automation also tend to raise the bar for reliability, safety checks and compute capacity.

Meta’s reach is the advantage — and the cost problem

Meta has a distribution engine few companies can match. Its “family daily active people” metric, which counts people who open at least one app in Meta’s ecosystem on a given day, rose 4% from a year earlier to 3.56 billion . At that scale, even a feature used by a fraction of users can become a major infrastructure load.

That is why investors are not only asking whether Meta can build an agentic assistant. They are asking how expensive it will be to run.

In late April, Meta raised its 2026 capital expenditure forecast to US$125 billion to US$145 billion, up from a previous forecast of US$115 billion to US$135 billion . Meta shares fell in extended trading after the update .

Investing.com has described Meta as being in a heavy AI buildout phase focused on expanding data center and GPU capacity, with spending pressure weighing on free cash flow and investor confidence in the near term . The Motley Fool has also warned that higher capital expenditure could pressure free cash flow and operating margins .

How Meta could make the spending case

The optimistic case is straightforward: if an agentic assistant makes Meta’s apps more useful, users may spend more time completing tasks inside the company’s ecosystem. The clearest reported path so far is the planned agentic shopping tool for Instagram .

Meta already uses AI algorithms across Facebook, Instagram, Messenger and WhatsApp, according to The Motley Fool . A more capable assistant would push that strategy further — from AI that ranks content or supports platform features toward AI that helps users take actions.

But the public evidence on direct monetization is still thin. A product in internal testing does not yet prove that users will rely on it daily, that businesses will pay for it, or that shopping and engagement gains will be large enough to absorb the infrastructure cost.

Three things to watch next

First: what the assistant is actually allowed to do. Will it mostly recommend and draft, or will it complete tasks inside Facebook, Instagram, WhatsApp and Messenger? Reports so far point to a personalized assistant for everyday tasks and a separate Instagram shopping tool .

Second: when Meta moves beyond internal testing. Staff testing suggests progress, but the reports do not yet establish a public launch date or the real-world level of autonomy users will see .

Third: spending discipline. With Meta’s 2026 capex forecast now at US$125 billion to US$145 billion, the company needs to show that AI infrastructure can translate into growth, efficiency or revenue — not just technical capability . Until then, Meta’s agentic assistant will remain both one of the most ambitious consumer AI projects in Big Tech and a major test of investor patience.