Meta Plans Data Center Rentals and Undercuts AI Model Pricing

Meta's stock climbed more than 5% Friday, turning positive for the year after CEO Mark Zuckerberg outlined two potential revenue streams designed to offset the company's substantial AI infrastructure spending.

In a Thursday interview with Bloomberg, Zuckerberg revealed Meta is exploring the possibility of renting its AI computing power to external customers. The move would represent a significant strategic shift for the social media company, which has historically focused on advertising revenue.

New Cloud Computing Play

While Zuckerberg did not specify implementation details, Meta could pursue two distinct approaches: operating like Amazon Web Services by hosting and powering third-party AI models for customers, or functioning as a neocloud provider that sells direct access to its AI chips and servers.

The timing aligns with Meta's announcement Wednesday of a 33rd data center, this one in Canada. AI companies have consistently reported resource constraints and insufficient computing capacity to meet customer demand, suggesting a viable market opportunity for Meta's excess infrastructure.

Aggressive AI Model Pricing

Meta also launched Muse Spark 1.1, a new AI model accompanied by pricing that dramatically undercuts established competitors. The company will charge developers $1.25 per million input tokens and $4.25 per million output tokens.

Those rates represent a substantial discount compared to Anthropic's Opus 4.8 model, which Meta positions as comparable to Spark 1.1. Anthropic currently charges $5 per million input tokens and $25 per million output tokens—making Meta's output pricing roughly 83% cheaper.

Tokens serve as the fundamental unit of measurement in AI models, with input tokens representing user queries or commands and output tokens comprising the model's responses. For context, a token typically represents pieces of words or short phrases.

Why It Matters

Meta's dual strategy addresses investor concerns that have weighed on the stock year-to-date: rising capital expenditures without clear returns on AI investments. By potentially monetizing existing infrastructure through cloud services and attracting price-conscious developers with competitively priced AI models, Meta is creating pathways to revenue that extend beyond its core advertising business. The data center rental approach, if executed, would position Meta as a direct competitor to established cloud providers in the rapidly growing AI infrastructure market.

The pricing strategy for Spark 1.1 targets developers who prioritize cost efficiency over the maximum capabilities offered by premium models from OpenAI and Anthropic. This could help Meta build developer mindshare and ecosystem adoption while its Meta Superintelligence Labs continues advancing more powerful models.

These details were first reported by Yahoo Finance.