Meta teases agentic shopping tools and a massive AI rollout

Big tech spent the last two years pouring billions into AI infrastructure. Now, Mark Zuckerberg thinks he knows exactly how to get that money back: by turning AI into your personal shopping agent.

On Meta’s quarterly earnings call on January 28, Zuckerberg told investors that users will begin seeing a wave of new AI models and products roll out in just "a matter of months." According to him, 2025 was all about rebuilding the foundations of Meta's AI program, and 2026 will be the year those tools hit the frontier.

While he stayed vague on exact launch dates, he made one focus crystal clear: agentic commerce.

The secret weapon is your social context

Zuckerberg teased new "agentic shopping tools" designed to browse business catalogs and pick out the exact right products for users.

Meta isn’t the first to eye transaction-capable AI. Both OpenAI and Google have already pushed out platforms for agentic commerce, partnering with heavyweights like Stripe and Uber to handle the actual mechanics of moving money and booking rides.

Zuckerberg, however, believes Meta holds an unfair advantage: your personal data.

He argued that what really makes an agent valuable is the context it can see. While rival chatbots might only know what you just typed into a prompt, Meta knows your interaction history, personal interests, shared content, and social relationships. That data layer, Meta claims, will deliver a uniquely tailored shopping experience.

The push follows Meta's acquisition of general-purpose agent developer Manus back in December. Meta noted at the time that it would keep selling Manus while baking its agent tech directly into core apps.

A $135B bet on "Superintelligence"

To power this vision, Meta is preparing to spend an eye-watering amount of cash.

The company announced its 2026 capital expenditures will land between $115 billion and $135 billion. That is a massive jump from $72 billion in 2025. In filings, Meta pinned the increase on its core infrastructure and its newly formed Meta Superintelligence Labs.

While wild, that number is still below the rumored $600 billion Zuckerberg reportedly projected for infrastructure spending through 2028.

Investors have previously pushed back on Meta for not clarifying how these massive AI checks will ever translate to the bottom line. Promising tangible shopping agents hitting consumer hands within months looks like an attempt to calm those nerves.

Why it matters

For consumers and businesses, this could be the first time AI agents feel genuinely practical rather than just novel. Instead of scrolling past sponsored ads on Instagram, an assistant that actually understands your taste could handle product discovery and checkout in a single turn. The obvious friction point, of course, is trust: how many users actually want Meta analyzing their social footprint to sell them things?

For the broader AI race, this signals a major pivot from foundation models to transaction engines. OpenAI and Google are building open rails with payment partners, but Meta already owns the social graph and a massive merchant catalog. That gives it an organic distribution loop others can only envy.

That said, Meta's strategy puts immense pressure on execution. Pushing capital spending to $135 billion means Wall Street will expect immediate commercial returns. If these upcoming agents don't move the needle on advertising or merchant revenue this year, skepticism over Meta's AI splurge will only grow louder.