Meta has transitioned from offering its AI models freely to charging developers for the first time.
Developers using Meta's new API, which allows them to integrate the AI into their applications, pay approximately one-quarter the price that OpenAI and Anthropic list for comparable offerings. CEO Mark Zuckerberg called the pricing "very aggressive and attractive" in an interview. He said other labs have "very high margins" and that Meta sees an opening to offer "high-level intelligence at a much more affordable cost."
This represents a notable departure for a firm that had previously championed open-source releases. Zuckerberg argued that charging for the API lets Meta control the technology: "If you really want to build the best experiences for people, you have to be able to shape the underlying technology."
Meta is investing hundreds of billions into data centers, AI chips, and supercomputers as part of its push toward superintelligence. The company's projected capital expenditures for 2026 are a record.
Get the market news that matters in a five-minute read with Market Briefs, our free daily newsletter
But investors are not fully sold. To address investor worries, Meta unveiled a consumer chatbot subscription, an AI agent for businesses, and a planned cloud service offering computing power and AI infrastructure. By undercutting rivals, Meta aims to capture market share in the fast-growing AI API market, forcing competitors to reconsider their own pricing models.
Meta's shift from free, open-source models to a paid API reflects a wider industry trend. Competitors like Google, Microsoft, and Amazon have long monetized their AI through cloud platforms and API fees. Meta, by contrast, built its early reputation on releasing models like LLaMA for free, hoping to attract developers and accelerate ecosystem growth.
Now, with escalating infrastructure costs - including a rebuilt AI lab led by former Scale AI executive Alexandr Wang - the company needs revenue to justify its enormous spending. The upcoming model codenamed "Watermelon" is expected to push intelligence further, and Zuckerberg has signaled that Meta will continue to balance open-source contributions with commercial offerings.
The new model tested better than Google's Gemini in several categories: agents, coding, and multimodal tasks. Zuckerberg called it a milestone: "This may be the first time, at least that I can remember, that Meta's models are better than all of the Google models."
Meta's ability to maintain a low price while still funding its massive AI infrastructure will be closely watched by investors. The $10 billion Canadian data center and continued heavy spending mean the company must convert developer interest into steady revenue. The upcoming "Watermelon" model and cloud service will test whether Meta can sustain its aggressive pricing without sacrificing margins or quality.
Join Market Briefs, our free daily newsletter, for a quick daily rundown of the markets