Meta Platforms (NASDAQ:META) took a significant step into the commercial AI space on Thursday by launching its first public AI model service for developers. The move, announced from Menlo Park, signals a strategic pivot that could reshape how investors view the social media giant's massive infrastructure investments.
Strong External Demand
CEO Mark Zuckerberg revealed in an interview with Bloomberg that external demand for Meta's computing power is so strong it could eventually rival the company's internal projects. "The offers that you get for using the compute are so high," Zuckerberg stated, though he emphasized that Meta is not sitting on idle capacity. The company's new service, called Muse Spark 1.1, is now available to U.S. developers through the Meta Model API at competitive pricing: $1.25 per million input tokens and $4.25 per million output tokens.
Monetization Potential
Shay Boloor, a strategist at Futurum Equities, described the API as a "much clearer monetization bridge" provided the model maintains strong coding performance. This comes at a critical time as Meta prepares to invest between $125 billion and $145 billion in capital expenditures in 2026, primarily for servers and data centers. That figure represents an 87% increase over the $72.2 billion spent in 2025 and accounts for 67% of Meta's $201 billion in revenue last year. Meta shares closed at $631.48 on Thursday, up 4.7% ahead of Friday's session.
Capacity and Chip Strategy
Internally, Meta is targeting 7 gigawatts of computing capacity this year and 14 gigawatts by 2027. The company also plans to begin production of its custom Iris AI chip in September. "You can't become an AI titan if you are dependent on another company for chips," said Forrester analyst Mike Gualtieri, noting that custom silicon can reduce model costs.
Revenue Projections
Analyst firm SemiAnalysis estimates that a 200-megawatt slice of Meta's capacity could generate more than $10 billion annually if sold at premium prices. That 200MW represents just 1.4% of Meta's planned 2027 capacity, about 5% of 2025 sales, and 16% of Meta's $62.8 billion capex increase. While a straight-line extrapolation would suggest over $700 billion—more than triple last year's revenue—this figure reflects scarcity pricing for a small block, not the entire fleet.
Cloud Division Comparison
For context, Amazon's AWS generated $37.6 billion in Q1 2026 with a 37.7% operating margin, while Alphabet's Google Cloud reported $20.0 billion with a 32.9% margin. Meta has yet to disclose any revenue or margin figures for its Model API preview, which remains in public beta.
Strategic Challenges
Meta faces a potential channel conflict, as it has agreed to spend up to $21 billion with CoreWeave (NASDAQ:CRWV) for compute capacity through 2032, according to a CoreWeave filing. This positions Meta as both a major customer and a potential selective rival. The risk is that scarcity premiums could vanish before Meta has supply ready, or that renting chips could divert focus from core areas like advertising or model training. Motley Fool writer Marc Guberti noted Meta is "several years away" from offering neocloud services, as it still relies on third-party capacity and major new sites take time to build.
Investor Focus
For investors, key metrics to watch in upcoming updates include how much capacity Meta allows outsiders to reserve, contract lengths, pricing, profitability after direct service costs, and whether customer deals reduce future capital spending. Meta doesn't need to replicate AWS to generate returns; even selling a small portion of its capacity can improve the economics of its massive buildout, as long as it doesn't hinder products powered by the rest.
