Meta’s “AI Shopping Agents” Are Coming in Months — And Zuckerberg Is Betting $135 Billion That You’ll Use Them

Meta CEO Mark Zuckerberg says the company will begin rolling out new AI models and products “in a matter of months,” marking 2026 as the year Meta shifts from building AI infrastructure to pushing consumer-facing AI tools across Facebook, Instagram, and WhatsApp.

The headline feature: agentic commerce — AI-powered shopping assistants designed to help people find “just the right set of products” from Meta’s business catalogue, potentially turning social feeds and chat threads into end-to-end shopping experiences.

Behind the strategy is a staggering infrastructure push. Meta said it expects $115–$135 billion in capital expenditures in 2026, up from $72 billion in 2025, explicitly tied to scaling its AI efforts under its restructured AI organization.

2025 Was the Build Year. 2026 Is the “Shipping” Year.

On Meta’s earnings call, Zuckerberg described 2025 as the year the company rebuilt the foundations of its AI program — a reorganization that culminated in a newly restructured AI lab and major investment in infrastructure.

Now, he’s signalling a pivot from internal capability-building to public releases.

“Over the coming months, we’re going to start shipping our new models and products,” he said, adding that Meta expects to “steadily push the frontier” over 2026.

Notably, Meta did not provide specific product names, hard launch dates, or technical benchmarks for its upcoming models — implying a phased rollout rather than a single “big reveal.”

What Is “Agentic Commerce” — And Why Meta Thinks It’s a Money Printer

Zuckerberg’s most concrete near-term use case was commerce. Meta’s vision is a shift from traditional shopping (search → browse → compare → buy) to AI agents that can handle discovery and decision-making inside Meta apps.

As he put it: “New agentic shopping tools will allow people to find just the right set of products from the businesses in our catalogue.”

In plain terms, these tools are expected to behave more like a shopping concierge than a search box. Instead of typing keywords and scrolling endless results, a user could describe what they want in normal language and receive a curated shortlist — potentially with follow-up Q&A, comparisons, and purchase flow handled inside the app.

Meta’s advantage is distribution: its apps reach billions of users, and over 200 million businesses already use its platforms in some form — meaning Meta can embed shopping agents exactly where people already spend time (feeds, Stories/Reels, DMs, and business chats).

Meta’s “Secret Weapon”: Personal Context at Massive Scale

Meta is also pitching something competitors can’t replicate easily: deep personal context.

Zuckerberg said Meta is “starting to see the promise of AI that understands our personal context,” including “our history, our interests, our content and our relationships,” arguing that the unique context Meta can see enables “a uniquely personal experience.”

That’s the core bet: context + distribution can beat pure technical advantage.

In practice, this “context” could mean AI agents that personalize recommendations using years of behaviour signals — what users watch, like, save, click, message about, and who they interact with. Done well, it could make shopping suggestions feel dramatically more relevant than generic e-commerce search.

But it also raises the obvious tension: the more personal the agent, the more questions users and regulators will ask about consent, transparency, and data boundaries, especially in strict regulatory regions.

Manus: Meta Didn’t Just Build Agents — It Bought Them

Meta is also accelerating by acquisition. TechCrunch reports that Meta acquired Manus, described as a general-purpose agent developer, and said it would both continue operating the service and integrate it into Meta products.

That matters because “agentic” systems aren’t just chatbots — they’re meant to execute multi-step workflows (search, filter, compare, transact). Buying agent infrastructure can shorten the time between “research demo” and “product people actually use.”

The Price Tag: $115–$135 Billion in 2026 Capex

Meta’s spending guidance is one of the biggest signals that it’s going all-in.

The company expects to spend $115–$135 billion on capital expenditures in 2026, up from $72 billion in 2025, and attributed the jump to “increased investment” supporting its AI lab efforts and core business.

That level of spend implies:

• more data centers,
• more AI accelerators and networking,
• and more capacity to train and serve models at global scale.

Investors appear cautiously receptive — Meta’s stock rose sharply after earnings, with reports highlighting market optimism even as some analysts flag the risk of heavy spending weighing on profitability.

The Real Question: Will People Actually Shop This Way?

Agentic commerce is a hot idea across the industry. TechCrunch notes that Google and Open AI are also building platforms for agent-enabled transactions, with partnerships that connect agents to payments and services.

And the broader retail world is actively testing shopping chatbots — with major interest, but also concerns about fraud, trust, and who controls the customer relationship when an AI becomes the “storefront.”

For Meta, success depends on whether users embrace:

• AI-driven product discovery inside social apps,
• AI-led shopping conversations in messaging,
• and the idea of an assistant that uses personal context to recommend purchases.

If adoption clicks, Meta could monetize through:

• transaction cuts,
• business tools and premium storefront features,
• and higher-performing ads driven by better targeting and conversion.

If it doesn’t, the risk is obvious: massive capex, lukewarm user behaviour change, and renewed investor skepticism about ROI.

Bottom Line

Meta is framing 2026 as the year it stops “building the engine” and starts driving the car — shipping new AI models and products within months, and pushing agentic commerce as the first major revenue-linked use case.

Zuckerberg’s bet is bold: AI agents that know you (context) inside apps you already use every day (distribution), backed by infrastructure spending on a scale few companies can match. Whether this becomes the next era of shopping — or just another AI feature users ignore — should become clear as rollouts begin in the first half of 2026.