Meta widens its enterprise AI bet beyond agents, Zuckerberg says

Meta is moving beyond a single enterprise AI product and aims to sell a broader mix of business tools, APIs, compute and services, CEO Mark Zuckerberg said on Wednesday’s second-quarter earnings call. The expansion matters because it signals a push to create new revenue streams outside advertising while leveraging Meta’s massive AI infrastructure and installed base of businesses already using its platforms.

The company only entered the enterprise AI market in June with an AI agent designed to help businesses handle customer service, support and other routine work. But Zuckerberg told investors that Meta sees a much larger opportunity ahead, including products that could be sold to both small businesses and bigger corporate customers.

That broader strategy comes as Big Tech companies race to monetize generative AI in ways that complement their existing businesses. For Meta, that means building tools that can plug into messaging apps, customer interactions, internal workflows and potentially even the sale of computing power to outside customers.

What Meta is trying to sell

Meta’s enterprise AI pitch now appears to stretch well past customer-service bots. Zuckerberg described a business opportunity that could include application programming interfaces, business-facing agents, internal productivity software, and in some cases compute infrastructure itself.

Those offerings would sit alongside Meta’s core advertising business and its smaller subscription revenue streams. In effect, the company is trying to turn its AI investments into a second growth engine, not just an efficiency tool for ads and social apps.

Why the existing advertiser base matters

The company’s first target is unsurprising: the millions of advertisers and small businesses already active across Meta’s platforms. Zuckerberg framed the new AI tools as a natural extension of Meta’s long-standing relationship with those customers.

In practice, that means AI agents could help businesses communicate with customers through WhatsApp, Messenger, Instagram and Facebook, while Meta potentially earns money when those tools deliver results. The model echoes the company’s performance-based advertising business, where businesses pay to reach users and see measurable outcomes.

Zuckerberg said Meta expects to be compensated in a way similar to its ad business, with payment tied to the results delivered for businesses using the tools.

That approach could make Meta’s AI products easier to sell to advertisers already accustomed to tracking leads, conversions and customer engagement across its apps.

How could Meta expand beyond small businesses?

Meta could broaden its reach by packaging internal software it built for itself and offering it to external customers later. Zuckerberg said the company is already developing coding tools, development tools and internal productivity systems for its own use, and those products may eventually become marketable enterprise software.

That would put Meta in a more traditional enterprise software lane, one it has not historically occupied in a major way. It would also open the door to larger organizations that want AI tools for employee workflows, software development or customer operations.

Still, Zuckerberg acknowledged that selling into the enterprise is a fundamentally different discipline than selling ads. He described it as a new capability Meta must learn rather than a simple extension of the company’s current strengths.

He noted that enterprise sales require a “different muscle” than the one Meta has relied on in its consumer and advertising businesses.

Why compute is now part of the conversation

One of the most notable parts of Zuckerberg’s comments was Meta’s interest in potentially selling compute to enterprise customers. Compute, in this context, refers to the expensive hardware and processing capacity that powers modern AI systems.

Meta has invested heavily in its own infrastructure to support large language models and other AI products. Zuckerberg said the company believes there are moments when it can sell that capacity externally at a “significant premium” to what it paid for it.

At the same time, he stressed that Meta cannot simply liquidate all of its computing power for short-term profit. The company needs to keep enough of that infrastructure for its own long-term AI ambitions, especially as it moves closer to what Zuckerberg called personal superintelligence.

Zuckerberg said it would be shortsighted to sell off all available compute, adding that Meta is managing the business as a portfolio of short-term and long-term priorities.

That balancing act is likely to remain central to Meta’s AI strategy. Every decision about how much infrastructure to sell versus reserve for in-house use affects both near-term revenue and the company’s ability to compete in the next wave of AI products.

What is Meta’s bigger AI vision?

Meta is not only thinking about enterprise software. The company is also betting that agentic AI — systems that can act on behalf of people or businesses rather than simply answer questions — will become a major platform shift across consumer and professional use cases.

That means Meta’s AI strategy spans both sides of the market. Businesses may get agents for support, sales and operations, while consumers are being promised personal AI agents and smart glasses that can interact with the world around them.

The consumer hardware angle is especially important because Zuckerberg sees future AI interactions as increasingly tied to devices that can blend digital and physical experiences. That could create demand for new kinds of hardware that are better suited to always-on, context-aware AI assistants.

How Meta is using AI inside its own apps

Meta is also applying large language models to speed up the development of its own social products. The company has recently launched several AI-assisted experiments, including tools for Marketplace sellers, Facebook Groups, and vibe-coded games.

Those launches suggest that Meta is using generative AI not only as a product for customers but also as a way to make app creation faster and cheaper internally. If that works, the company could ship more features, test more product ideas and scale promising ones more quickly through its recommendation systems.

Zuckerberg said he expects product development to become significantly easier as AI tools become more integrated into Meta’s workflow. That could give the company an advantage in rapidly turning experiments into full-scale products.

He said Meta plans to generate more ideas and then use its recommendation systems to distribute the most promising ones to users likely to care about them.

Key areas in Meta’s enterprise AI strategy

The company’s comments point to a multi-pronged approach rather than a single product line. The table below summarizes the main pieces of the strategy described on the earnings call.

Why this matters for Meta’s business model

Meta remains overwhelmingly dependent on advertising, and that is unlikely to change overnight. But the company has been clear that it wants AI to do more than improve ad targeting or content recommendations. It wants AI itself to become a product category that can generate direct revenue.

That is important for two reasons. First, it diversifies the company’s business at a time when digital advertising can be cyclical and competitive. Second, it gives Meta a way to capture more value from the infrastructure spending that has already gone into training and running large AI systems.

If successful, the strategy could help Meta move from being primarily a consumer internet company to something closer to an AI platform provider with multiple lines of business. The challenge is execution: enterprise buyers demand reliability, support, security and integration, all of which require a very different operating model from social media.

What risks does Meta face?

The biggest risk is that Meta may be trying to enter too many adjacent markets at once. Enterprise software, AI infrastructure and consumer AI hardware each require different sales motions, product expectations and technical support.

Another risk is cannibalization. If Meta sells too much compute externally, it could limit what it has available for its own AI roadmap. If it sells too little, it may leave money on the table in a fast-growing market.

There is also the broader question of whether businesses will trust Meta as a serious enterprise vendor. The company has enormous scale and strong technical capabilities, but it will need to convince corporate customers that it can deliver durable, secure and useful AI products over time.

Timeline: how Meta’s enterprise AI push is unfolding

Meta’s AI move has developed quickly over the past two months, and the pace suggests the company is still defining the exact shape of its business strategy.

What comes next?

Meta’s next step will likely be proving that it can translate its AI ambition into products businesses actually want to buy. The company has already identified a large existing customer base through its advertiser network, which gives it a distribution advantage many AI startups do not have.

But enterprise AI is a crowded field, with major competitors offering their own agent platforms, developer tools and cloud infrastructure. Meta will have to show that its products are not just technically capable, but also easy to deploy, reliable in production and worth paying for.

If it succeeds, the payoff could be substantial. Meta would not only deepen its AI footprint across consumer and business markets, but also create new income streams that are less dependent on ad cycles and more directly tied to software, services and infrastructure.

For now, Zuckerberg’s message to investors was clear: the enterprise opportunity is bigger than one agent, and Meta intends to explore the full stack.

Key takeaways

• Meta says its enterprise AI plans go well beyond the business agent it launched in June.
• The company is considering APIs, business agents, compute sales and internal tools as future products.
• Meta wants to monetize AI through its advertiser base and potentially larger enterprise customers.
• Zuckerberg warned that selling compute requires balancing short-term revenue against long-term AI needs.
• Meta is also using AI to speed up development of its own apps and product experiments.

FAQ

What is Meta’s enterprise AI strategy?

Meta’s enterprise AI strategy is to sell a wider set of business tools than its current AI agent, including APIs, AI services, internal productivity software and possibly compute capacity. The company wants to turn its AI investments into a new revenue stream that complements advertising.

Why is Meta interested in selling compute?

Meta is interested in selling compute because it believes demand for AI infrastructure is strong enough to support premium pricing. The company says it may be able to earn more from outside customers, but it must keep enough capacity for its own long-term AI plans.

Who is Meta targeting first with its AI products?

Meta is targeting its existing base of advertisers and small businesses first. Those customers already use Meta’s platforms to reach users, so AI agents that help with customer interactions and daily operations could be an easier sell than completely new enterprise software.

How is Meta using AI inside its own products?

Meta is using large language models to accelerate the development of internal and external products, including tools for Marketplace sellers, Facebook Groups and games. The company says AI should make it faster and easier to launch new apps and scale promising ideas.

What are the main risks for Meta’s enterprise AI push?

The main risks are execution, competition and infrastructure trade-offs. Meta must learn enterprise sales, compete against established software and cloud vendors, and decide how much compute to reserve for its own AI roadmap versus selling to others.