Meta AI Unveils Dual Breakthroughs: Launches First Paid AI Model and Mass Production of Iris Chip Imminent, Stock Surges 4.7%

Multiple foreign media outlets reported that Meta launched its artificial intelligence (AI) model Muse Spark 1.1 on Thursday (9th), officially entering the AI programming and Agentic AI market. The company also unveiled its first API pricing plan, aiming to compete directly with industry leaders like OpenAI and Anthropic through more competitive pricing, while accelerating the monetization of its AI business.

Meta's stock surged 4.7% on Thursday, closing at $631.48. The market welcomed the launch of Muse Spark 1.1, Meta's first paid AI programming model, along with news that its in-house AI chip Iris is set for mass production in September. These developments are expected to enhance AI monetization, reduce data center costs, and alleviate investor concerns over massive capital expenditures.

On Thursday, Meta officially released Muse Spark 1.1, an Agentic AI programming model whose performance rivals that of leading players like OpenAI and Anthropic, marking Meta’s formal entry into the enterprise AI programming market.

Nick Jones, an analyst at BNP Paribas, noted that Muse Spark 1.1 is Meta’s first AI model offered on a paid basis, representing the company’s first step toward generating direct revenue from AI products. Developers can access the model via the newly launched Meta Model API platform, marking a new milestone in Meta’s AI commercialization journey.

Jones added that if Meta further launches cloud computing services alongside its AI model API pricing, it could generate new revenue streams beyond advertising, indicating that its AI investments are beginning to establish a clear path to profitability.

In recent years, Meta has heavily invested in AI infrastructure, including establishing the Meta Superintelligence Labs, recruiting top AI talent, and building large-scale data centers. However, the market has consistently questioned when these multi-billion-dollar investments would yield tangible returns.

Earlier this month, reports emerged that Meta plans to offer data center computing power to third parties through its newly formed Meta Compute division, sparking mixed analyst reactions. Some believe this reflects a strategic shift to control capital spending, while others see it as a move to open new revenue channels.

Jones emphasized that both future cloud services and the Muse Spark API monetization model will generate incremental revenue and “demonstrate a clear and immediate high-return path for Meta’s AI investments.”

On the same day, Reuters reported that Meta’s self-developed AI chip, codenamed Iris, is expected to begin mass production in September—a significant milestone since the launch of the Meta Training and Inference Accelerator (MTIA) program in 2023.

The market believes that if Iris achieves successful mass production, it could reduce Meta’s reliance on NVIDIA (NVDA-US) AI chips, lower AI computing and data center deployment costs, and further enhance the profitability of Meta’s AI operations.

Reuters also noted that Meta plans to double its data center computing capacity from 7GW in 2026 to 14GW in 2027. This news initially raised concerns about rising AI capital expenditures, putting downward pressure on Meta’s stock during early trading. However, the release of Muse Spark 1.1 and the Iris mass production news boosted market confidence, leading the stock to close sharply higher by 4.7%.

Brent Thill, an analyst at Jefferies, stated that Meta will continue to expand its AI investments and cloud infrastructure. Meta recently revealed that its Prometheus data center in Ohio, USA, went live in July. The company expects that as AI demand continues to grow, demand for computing power and infrastructure will further increase.

Benjamin Black, an analyst at Deutsche Bank, estimated that if Meta adopts a hybrid deployment of Iris and NVIDIA chips in its data centers, data center costs could be reduced by up to 35% by 2027.

Black added that the ongoing progress of the MTIA program will help improve the cost structure for AI inference and recommendation systems, suggesting that Meta’s incremental costs may not be as high as previously feared. Meanwhile, the high-margin revenue opportunities from AI could exceed market expectations.