In June, Meta entered the enterprise AI market with a dedicated AI agent for businesses, designed to handle customer service, support, and daily operations. But the company’s ambitions now appear much larger. During Meta’s second-quarter earnings call on Wednesday, CEO Mark Zuckerberg told investors that the enterprise opportunity extends well beyond those initial agents, spanning APIs, direct compute sales, and other services for large customers.
“We see a large enterprise opportunity to sell to businesses, including APIs, business agents, potentially selling compute directly, and other services that we’re building for large customers,” Zuckerberg said. These new offerings could eventually give Meta additional revenue streams beyond advertising, which still drives the vast majority of its overall business, and subscriptions, which contribute a much smaller share.
For now, Meta is expected to start with its existing base of advertisers. The company plans to offer AI agents that operate across messaging apps and other surfaces, enabling businesses to interact with their own customers through an AI interface. Zuckerberg emphasized that Meta will get paid when those agents deliver results, similar to how its ad system works today. He described the opportunity as an extension of the sales and partnership relationships Meta already has with millions of advertisers and hundreds of millions of small businesses using its platforms.
From small businesses to larger enterprises
Meta’s current advertiser base is heavily weighted toward small and medium-sized businesses. That means the first phase of the enterprise AI push will likely focus on those customers, who need low-cost, easy-to-deploy tools for customer service, lead generation, and sales support. But Zuckerberg said Meta is also building for larger organizations.
“There are other enterprise customers who I think we’re increasingly going to serve, too,” he explained. “We’re building coding and developing and internal productivity tools partially because we need to build them ourselves, and we need to make sure that we have tools that are tuned for ourselves. Now that we have those, we feel like there’s a large opportunity to serve — whether that’s small businesses or larger businesses.”
That shift, however, may not be easy. Zuckerberg admitted that selling to enterprise customers is “a different muscle” than the one Meta has historically relied on. For years, the company has focused on consumer apps, advertising technology, and developer platforms, but it has not built a large enterprise sales force comparable to legacy software companies. Building that capability would require new go-to-market strategies, support organizations, and product packaging, according to analysts who follow the company. Meta has made some strides with Workplace, its enterprise communication product, but the broader enterprise business remains nascent.
Compute as a potential revenue stream
One of the more surprising areas of opportunity Zuckerberg highlighted was direct compute sales. Meta has been investing heavily in data centers, custom silicon, and AI infrastructure, and the company believes it can sell excess capacity to external customers at a premium.
“We currently have the opportunity to sell compute at a significant premium over what we paid for it,” the company said on the call. But Zuckerberg cautioned against cashing in too aggressively. He told investors that it “would be foolish” to sell all of the compute and take a short-term profit. Instead, he described Meta’s approach as a “portfolio” that includes a mix of long-term and short-term plans for its compute infrastructure.
The reasoning is tied to Meta’s long-range goal of building what Zuckerberg calls “personal superintelligence.” As AI systems become more powerful and more personalized, they will require hardware that allows users to interact with them seamlessly. If Meta sells all of its compute now, it could sacrifice its ability to build those future products and maintain a competitive edge. The company therefore wants to keep enough capacity for its own research and product development while monetizing unused infrastructure when possible.
Agentic AI for consumers, too
Meta’s agentic AI plans are not limited to businesses. The company has promised to bring “personal AI agents” to consumers, alongside AI smart glasses that can understand and interact with the world in front of them. These agents are intended to perform tasks on behalf of a person, rather than simply answer questions. That could mean booking reservations, managing schedules, making purchases, or coordinating with other AI agents.
The consumer angle is important for Meta because its scale in social media gives it a massive distribution advantage. WhatsApp, Messenger, Instagram, and Facebook reach billions of people globally. By embedding agentic AI into these apps, Meta could introduce a new generation of AI-powered interactions without requiring users to download separate apps. It also gives the company more data and feedback loops to improve its models.
At the same time, Meta is using large language models to accelerate development across its social apps. Recent launches include an app for Marketplace sellers, another for Facebook Groups, one for vibe-coded games, and other experiments. Zuckerberg teased that more are on the way.
“I expect it to become a lot easier to ship new apps,” he said. “So we are planning to build out more ideas and use our recommendation systems to scale them to the people who will find them interesting.”
This approach reflects a broader trend: generative AI is lowering the cost of building software. Social platforms can now spin up niche apps in days or weeks rather than months. Meta’s recommendation systems, which already determine what billions of users see in their feeds, could then determine which new apps gain traction. That could allow Meta to expand its product family without making huge bets on a single app.
Balancing short-term gains and long-term vision
Meta’s earnings call made clear that the company sees enterprise AI as a multi-year opportunity. The initial focus on advertisers gives it a revenue path that is relatively easy to understand: businesses pay Meta when AI agents deliver measurable outcomes, much like they already pay for ads when they see a return on spend. Expanding to larger enterprises with internal tools and other services would be a bigger transformation.
The compute business represents another potential revenue stream, but it comes with a strategic warning. Meta has already committed enormous sums to AI infrastructure, and investors are watching to see when those investments will pay off. Selling compute at a premium could produce near-term upside, but Zuckerberg’s “portfolio” framing suggests the company will not sacrifice long-term projects for immediate profits.
The scope of what Meta is describing is unusually broad. Few companies can credibly claim to offer AI agents, enterprise APIs, direct compute sales, consumer smart glasses, and a rapidly expanding portfolio of social apps. But Meta’s scale, data, and distribution make at least part of that vision plausible. How much of it becomes real revenue will depend on whether Meta can execute in enterprise categories where it has limited experience.
For now, the company is signaling that it wants to be the infrastructure and application layer for the next wave of AI-powered business communication, just as it became the advertising backbone for millions of businesses in the previous internet era. Whether that opportunity is as large as the ad business remains unclear, but Zuckerberg is making clear that Meta intends to try.
Source: TechCrunch News