Zuckerberg says Meta’s enterprise AI opportunity extends beyond agents

Meta's earnings call on Wednesday started as the usual advertising-and-engagement recap, but Zuckerberg spent a good chunk of it talking about something that sounds more like Amazon Web Services than Instagram. He wants Meta to sell AI infrastructure and services directly to businesses — APIs, business agents, and, notably, compute itself. That's a meaningfully different pitch than "we'll show your ad to more people," and Zuckerberg knows it. He called enterprise sales a "different muscle" for a company that has spent two decades getting extremely good at one thing: advertising.

The near-term plan leans on what Meta already has going for it — millions of small businesses and advertisers already embedded in its platforms. The company launched a customer-service-focused AI agent back in June, and Zuckerberg's framing on the call made clear this is meant to work like the ad business itself: Meta gets paid when the agent actually delivers results for a business, not just for existing. Think of it as performance marketing, except the product is a chatbot handling support tickets instead of a banner ad.

Where it gets more interesting is the compute question. Meta has built out a mountain of AI infrastructure to train and run its own models, and Zuckerberg openly acknowledged the company could sell that capacity to outside customers at a serious markup over cost. But he was equally blunt that cashing in on that opportunity wholesale would be a mistake — his word was "foolish." Instead, Meta is treating its compute like an investment portfolio, keeping a mix of long-term and short-term allocations rather than liquidating capacity for quick profit. That's a notably disciplined stance from a company that just spent Q2 explaining to investors why its AI spending keeps climbing.

There's a second, quieter angle here too: internal tools becoming external products. Meta has been building coding assistants and productivity software for its own engineers, largely out of necessity, and Zuckerberg suggested those same tools could eventually be offered to outside businesses, small and large alike. Pair that with his comments about large language models letting Meta ship new consumer apps faster — Marketplace tools, a Facebook Groups app, AI-assisted game-making — and the picture is a company trying to turn every internal AI investment into multiple external revenue lines simultaneously, rather than betting everything on one.

None of this replaces advertising as Meta's core business anytime soon; Zuckerberg was careful to frame enterprise AI as additive, not transformative, at least for now. But the ambition on display — agents, APIs, compute-for-rent, and consumer superintelligence hardware all mentioned in the same call — shows a company no longer content to be judged purely on engagement metrics and ad prices.