Meta and Wall Street’s new debate: could AI compute become a business beyond ads?

Meta Platforms is drawing fresh attention for the possibility of turning its AI build-out into a broader infrastructure business, a shift that could reshape how investors think about the company’s revenue streams, not just its technology spending.

The renewed discussion was sparked by remarks from CNBC host Jim Cramer and by an accompanying JPMorgan view cited in market coverage. Cramer said Meta is getting into AI’s “most lucrative game,” referring to a strategy described in reports as business-to-business sales of access to AI models hosted on Meta infrastructure, alongside renting excess computing capacity.

Coverage also points to the idea that Meta could charge developers to use AI models that run on its systems. That model resembles the way other AI cloud providers lease GPU clusters and related compute resources to enterprise customers. In this framing, the monetization question is less about ads served through social apps and more about compute-as-a-service and platform-like distribution for AI builders.

JPMorgan’s analysis, as reported, argues the opportunity could be materially larger than consensus expectations. One figure cited in the market commentary was a potential $20 billion business tied to AI compute and access, though neither JPMorgan nor Meta was described as having publicly confirmed a specific commercial launch, pricing schedule, or timetable.

Cramer’s comments included market reaction context. In the same discussion, he referenced a stock move that was “up only 49 points” while he suggested the company was pursuing a higher-value commercial direction. The commentary also referenced an “18x EPS” comparison, indicating that the debate is not only about product strategy but about whether Meta’s market valuation adequately reflects possible AI-driven revenue growth.

For Meta, the attraction of an AI infrastructure business is straightforward even without detailed disclosures. The company has invested heavily in AI research and training, and the costs of acquiring and operating high-performance computing systems are significant. Turning parts of that capacity into billable services could, in theory, create a secondary revenue stream that grows with developer demand for model access and compute capacity, rather than being limited to ad budgets and engagement trends.

Still, key details remain unaddressed in the public discussion being circulated. As described in the market commentary, the concept rests on reported exploration of developer access and compute rental, but the information does not include contract structure, expected margins, geographic rollout, model availability, or how Meta would integrate such offerings with its existing developer tools and cloud partnerships. Without direct company disclosures, it is not possible to determine whether this is an internal feasibility effort, a limited pilot, or a broader commercialization plan.

More broadly, the sector context is that AI infrastructure is increasingly treated as a competitive battleground. Compute shortages and the economics of GPU usage have encouraged companies to look for ways to monetize excess capacity, while developers seek reliable and cost-effective access to models. If Meta were to follow through, it would join the ranks of providers positioning themselves as destinations for AI deployment, not only creators of AI research and social features.

What to watch next is whether Meta provides clearer confirmation, such as product naming, developer access terms, or any mention of infrastructure monetization in investor communications. Investors may also look for indicates in Meta’s spending, partnerships, or platform roadmaps, as well as for any public updates from regulators or counterparties that would clarify what, if anything, Meta intends to offer commercially.