Meta has launched a new business line called Meta Enterprise Platform, a move aimed at turning its AI capabilities into products sold directly to companies. Earlier this week, Chief Executive Officer Mark Zuckerberg wrote on X that Meta is starting the “next major pillar” of its business to help enterprises grow and transform with AI.
Meta’s newsroom listed the initial products as Muse agent, Meta Business Agent, Muse API, and Muse Code. The new unit is led by CJ Desai, who was recently recruited from MongoDB and reports directly to Zuckerberg.
Advertising still accounts for nearly all revenue
According to Meta’s 10-Q quarterly filing, the company generated $60.801 billion in revenue in the second quarter. Of that total, $59.363 billion came from advertising, equal to 97.6%. For the first half of the year, the share was 97.7%.
The filing also showed that second-quarter revenue rose 28% year over year, while advertising revenue increased 27%. Ad impressions, however, were up 14%, and the average price per ad rose 12%. That means growth came mainly from showing more ads to users and charging more for each ad, rather than from reaching a larger audience.
The report said Meta’s advertising engine is not broken, but relying on volume increases and higher pricing alone leaves limits to future growth, while AI spending is already climbing.
AI spending has hit margins and free cash flow
Meta’s capital expenditures in recent years have gone largely to servers and data centers. The company’s guidance for full-year 2026 capital expenditures is $130 billion to $145 billion.
Against that backdrop, total costs and expenses rose 55% year over year to $42 billion. Operating margin fell from 43.0% to 30.9%, net income declined 14%, and quarterly free cash flow dropped to $784 million, down 91% from a year earlier.
Meta’s financing posture has also shifted. The company reported zero share repurchases in the first half of the year, compared with $23.159 billion in the same period last year. In May, it also issued $25 billion of senior unsecured notes. The report argued that for a company with historically strong cash generation, halting buybacks while raising debt at this scale shows AI has become an increasingly heavy bill.
Turning compute from a cost center into revenue
The logic behind Meta Enterprise Platform is straightforward: turn compute from a cost center into a source of revenue. Until now, Meta has mainly recouped AI investment indirectly, through better ad targeting and higher ad pricing. Selling AI tools to businesses would create a direct charging model.
Zuckerberg said in his post that Meta serves billions of people and helps hundreds of millions of businesses reach customers. The report said that gives Meta something many rivals do not have: an existing distribution channel. Many businesses already interact with customers through Meta’s apps, so the company does not need to start from cold outreach when pitching new products.
In practical terms, the same pool of compute that once mainly improved ad performance could now support invoices sent directly to customers.
Business Agent already uses subscription and usage-based billing
Meta is not starting from zero on monetization. On June 3, Zuckerberg introduced Meta Business Agent at a company event in London. The product can answer customer questions, recommend products, and handle bookings across three messaging platforms. It is included in the business offering under the Meta One subscription.
For large enterprise clients on the WhatsApp Business Platform, Meta uses a usage-based model similar to message-based billing.
Zuckerberg said at the time that a clothing shop in Birmingham or a bakery in São Paulo could offer the same around-the-clock, highly personalized experience as a major brand. As models improve, he said, agents will take on more tasks and eventually help business owners run the whole operation.
The product was tested for free last October under the name Business AI in markets including Mexico and India.
The report described this as the same path now being expanded: Business Agent sits partly inside a subscription bundle and partly under usage-based pricing for larger clients, while Enterprise Platform scales that approach by packaging models, agents, and infrastructure for businesses.
Pricing, customers, and segment disclosure remain unclear
Meta has not provided several details yet. Its press release did not disclose pricing, and it did not name any customers.
In the announcement, Desai said AI will fundamentally redefine how organizations of all sizes innovate, grow, serve customers, and operate over the coming years. The report said the vision is broad, but for now it remains a vision.
Three questions stand out next:
- How Meta will price the offering.
- Who the first customers will be.
- Whether the company will break out the business as a separate line in its financial reporting so outsiders can see whether it makes money.
Competition is entrenched, and trust is still a hurdle
Meta is entering a crowded field. Enterprise AI and cloud services are already dominated by Microsoft, Google, and Amazon, while model providers such as OpenAI and Anthropic also sell directly to businesses.
The report noted that Meta’s access to hundreds of millions of businesses is a ready-made channel, but not a ready-made order book. A merchant that buys ads on Instagram may not be willing to hand customer service and sales workflows over to a Meta agent.
Trust is another issue. If companies are going to let agents handle customer conversations and transaction data, they first need to assess whether the vendor can protect that information. The report also said Meta’s record on privacy and child safety will be part of that review before enterprise clients sign contracts.