Meta and AI: Can it monetize anything beyond ads?

99% of Meta's revenue still comes from advertising. That's significant: it's the same pattern that has defined the company for over a decade, despite attempts to sell hardware (Portal, Quest), paid services (Workplace), and even a cryptocurrency (Libra/Diem). None gained meaningful scale. Now Meta is trying again, this time with artificial intelligence, and CNBC raises the question many in the industry are quietly asking: will this be different?

The honest answer, today, is that we don't know. But the ingredients of this bet are different from previous attempts, and it's worth examining them.

An AI strategy that doesn't depend on a single product

Unlike the metaverse, where Meta bet everything on a single shift in consumer behavior, its AI strategy is more distributed. On one hand, Llama (its family of open-weight models) has become the reference for companies wanting to deploy their own models without depending on closed APIs. On the other, Meta AI is integrated into WhatsApp, Instagram, Facebook, and Messenger, with an active user base exceeding 3 billion people. It's not a new product people need to be convinced to download: it's where people already are.

That solves the distribution problem, which is precisely what killed Portal and Workplace: no one wanted to go find them. But distribution is not the same as willingness to pay.

The structural problem: users trained not to pay

Meta has built its business on a simple premise: you use the product for free and we show you ads. That premise has worked extraordinarily well for generating advertising revenue, but it has created a user base with zero tolerance for paying for Meta services. WhatsApp tried charging a dollar a year over a decade ago and abandoned it. Meta Verified exists, but its adoption numbers aren't what the company expected.

Generative AI could break that pattern only if it offers enough differentiated value to justify a price. Models like OpenAI's ChatGPT Plus or Anthropic's Claude Pro have shown there's a user segment willing to pay for advanced capabilities. The question is whether Meta AI has or can develop those capabilities, and whether the Meta brand inspires enough trust for that kind of paid relationship.

The enterprise route: more realistic in the short term

Where Meta has better actual chances of monetizing AI without relying on changing consumer behavior is in the B2B segment. Llama is already used in production at thousands of companies. Meta could build on that: enterprise support, managed fine-tuning, dedicated infrastructure. It's a market where AWS, Azure, and Google Cloud have the advantage, but where Meta's open-weights bet gives it a differentiating angle.

The problem is that business requires an enterprise sales organization, relationships with CIOs, and SLAs that Meta historically hasn't had to build because its customer has always been the advertiser, not the tech lead at a mid-sized company.

Why this matters beyond Meta

This story isn't just about Meta. It's an indicator of whether large language models can sustain profitable businesses outside advertising and outside the tech developer segment. If Meta, with the world's largest distribution base, can't monetize AI significantly over the next two or three years, that says something important about the structural difficulty of the problem.

For the ecosystem of tools and agents built on third-party APIs—including those orbiting Claude and Anthropic's platform—the ability of major platforms to monetize AI directly affects how much capital continues flowing into the sector and how quickly the infrastructures on which we all build develop.

---

From our perspective, the reading is pragmatic: Meta has the distribution assets any company would envy, but changing the payment habits of its users is a problem of another magnitude. Whether it succeeds will largely determine what AI business model proves viable at massive scale.