July 10, 2026 — After three years of silence, Mark Zuckerberg returned to the X platform and dropped a bombshell that shook the entire AI industry: Meta officially launched the Muse Spark 1.1 multimodal reasoning model and simultaneously opened the public preview of the Meta Model API. This post signaled the official start of Meta’s transformation from an "AI technology provider" to an "AI infrastructure service provider."
This wasn’t just a routine product upgrade. Zuckerberg chose to make this announcement on X—his competitor’s turf—instead of Meta’s own social media ecosystem, sending a strong message in itself. One company is trying to unlock commercial potential with low-cost models, while another is searching for direction amid improving macro liquidity.
However, the capital market’s response was measured. As of July 10 (UTC+8), Meta’s stock closed at $631.48, up 4.70% for the day. For a tech giant, a 4.7% increase is far from lackluster, but compared to the "explosive effect" typically expected from an AI launch, the market’s enthusiasm was clearly restrained. Investors are no longer asking, "Does Meta have AI?" but rather, "Can AI make money?"
From Open Source to Paid: Why Meta Is Pivoting Now
To truly understand the significance of Meta’s latest strategic shift, we need to look back at the evolution of its AI journey.
Over the past two years, Meta’s AI strategy has revolved around "open source." From the successive open-sourcing of the Llama model series to building an AI research community, Meta aimed to gain developer trust and industry influence through an open ecosystem. Yet this approach faced a fundamental problem: open source doesn’t directly translate into revenue.
After a disappointing model release in spring 2025, Zuckerberg personally stepped in to rebuild the AI team, hiring Scale AI founder Alexandr Wang to lead the newly established Meta Superintelligence Labs. The company’s strategy gradually shifted from "open source first" to developing "proprietary, monetizable models." Muse Spark 1.1 is the first tangible result of this new direction.
Meanwhile, Meta’s infrastructure investment has reached staggering levels. In 2023, the company’s capital expenditures totaled $28.1 billion, rising to $39.2 billion in 2024, and hitting $72.2 billion in 2025. For 2026, Meta is set to ramp up annual capital spending to between $125 billion and $145 billion, with a focus on AI compute clusters and large model development—about double the 2025 investment. In just the first half of 2026, Meta signed contracts for over 5 GW of cloud computing and managed data center resources.
Such massive infrastructure spending demands a clear path to commercialization. The launch of Muse Spark 1.1 and the Meta Model API is essentially Meta’s way of creating a "revenue recovery channel" for these hundreds of billions in capital expenditures.
Muse Spark 1.1’s Differentiation: Low Price Doesn’t Mean Low Performance
From a product perspective, Muse Spark 1.1 is far from a hastily assembled response to market trends. According to Meta, this model is purpose-built for agent tasks, with significant enhancements in tool use, computer operation, code generation, and multimodal understanding. The model supports a 1 million token context window and can serve as a lead agent in multi-agent systems or as a specialized sub-agent. Zuckerberg revealed that Muse Spark 1.1 has outperformed Google’s Gemini model in several benchmarks, including agent capabilities, programming, and multimodal tasks.
What really caught the industry’s attention, however, is Meta’s pricing strategy. The Meta Model API is priced at $1.25 per million input tokens and $4.25 per million output tokens. Zuckerberg stated on X that this is roughly one-quarter the official price of comparable top-tier models from OpenAI and Anthropic. Registered developers also receive $20 in free credits to try out the service.
It’s important to note that this isn’t the "absolute lowest" price. It’s higher than OpenAI’s entry-level GPT-5 mini and Anthropic’s budget-focused Claude Haiku 4.5, but significantly lower than Anthropic’s high-end Claude Sonnet 4.6. Meta’s pricing targets the upper-mid-tier developer market—those who need robust model capabilities but are sensitive to the flagship pricing of OpenAI and Anthropic.
Four Giants, Four Strategies
Comparing Meta with OpenAI, Anthropic, and Google reveals four distinctly different commercialization logics.
OpenAI follows a "performance premium" model. Leveraging the technical lead of its GPT series, OpenAI charges enterprise clients high API fees and distributes model capabilities through Microsoft’s cloud channels. Its core assumption: as long as the model is strong enough, enterprises will pay a premium for performance.
Anthropic bets on a "safety premium." With "Constitutional AI" and safety as its differentiators, Anthropic attracts a large base of enterprise clients with strict compliance and risk control needs. Its secondary market valuation has soared to $1.2 trillion, reflecting the capital market’s recognition of the commercial value of "safe AI."
Google pursues a "full ecosystem integration" strategy. The Gemini model is embedded across Google’s entire product suite—search, ads, cloud, Workspace—where AI capabilities serve to boost ARPU from existing businesses rather than as a standalone revenue stream.
Meta, meanwhile, has chosen a fourth path: open ecosystem + cost advantage. By offering API pricing far below competitors, Meta aims to attract developers at scale, using ecosystem size to counter OpenAI’s technical moat and Google’s ecosystem moat. The logic is: lower prices → more developer adoption → larger ecosystem → data flywheel and network effects → long-term competitive advantage.
None of these paths is inherently superior, but Meta’s strategy stands out: it doesn’t rely on winning through a technical gap, but instead seeks to reshape the competitive landscape with an economic model. If the performance gap between AI models continues to narrow over the next 12–24 months, price will become a more decisive factor in enterprise decision-making—Meta’s core bet.
Why the Market Isn’t "All In"
After the announcement, Meta’s stock closed up 4.7% at $631.48. This would be impressive for any ordinary product launch, but considering Muse Spark 1.1 is Meta’s first enterprise-grade, revenue-generating model, the market’s reaction can best be described as "cautiously optimistic."
Investors aren’t doubting Meta’s AI capabilities; they’re concerned with three deeper issues.
First, the certainty of revenue contribution. With API pricing at just a quarter of competitors’, Meta needs to achieve several times the call volume of its rivals to reach comparable revenue. Muse Spark 1.1 is currently only available in public preview to US developers. There’s still a long road from preview to large-scale commercial adoption and meaningful revenue contribution.
Second, the sustainability of capital expenditures. Annual capital spending of $125–145 billion means Meta is burning over $340 million per day on AI infrastructure. Even if Meta’s ad business continues to grow—WARC Media forecasts $240 billion in ad revenue for 2026—such massive investment will keep pressure on the bottom line.
Third, the profitability timeline. AI infrastructure investments take time to generate profit. Goldman Sachs forecasts that the combined 2026 capital expenditures of Alphabet, Amazon, Microsoft, and Meta will reach $725 billion. With such enormous industry-wide investment, AI commercialization won’t be a story that plays out in just a quarter or two.
The market has moved from the "AI narrative" phase to the "AI delivery phase." Investors are no longer paying for "model launches"; they want to see how models convert into cash flow.
Conclusion
On the day Zuckerberg returned to X, Meta sent a clear message to the industry with Muse Spark 1.1 and the Model API: the AI race is shifting from "who has the best model" to "who can deliver models to the most people at the lowest cost."
OpenAI has a technical moat, Google has an ecosystem moat, Anthropic has a safety moat—Meta is betting on a price moat to move the market. Whether this strategy succeeds depends on two key factors: whether the performance gap between models is truly narrowing, and whether developers will actually switch for lower prices.
For the crypto industry, regardless of how this competition plays out, lower-cost AI infrastructure means more possibilities. When model calls are no longer a cost bottleneck, the imagination for on-chain intelligent applications will be redefined.
The story of AI commercialization is just turning to its second chapter. Chapter one was "who can build a model"; chapter two is "who can make models affordable and accessible." Meta is going all in on writing the second chapter.
FAQ
Q1: What is the exact pricing for the Meta Model API, and how does it compare to competitors?
The Meta Model API is priced at $1.25 per million input tokens and $4.25 per million output tokens. Zuckerberg says this is about one-quarter the official price of top-tier models from OpenAI and Anthropic. Registered developers also get $20 in free trial credits.
Q2: What are the core capabilities of Muse Spark 1.1?
Muse Spark 1.1 is a multimodal reasoning model purpose-built for agent tasks, with significant enhancements in tool use, computer operation, code generation, and multimodal understanding. The model supports a 1 million token context window and can serve as a lead agent in multi-agent systems or as a specialized sub-agent for specific tasks.
Q3: Why is Meta shifting from open-source Llama to a paid API model?
Meta’s annual AI infrastructure investment has reached $125–145 billion, and the open-source model can’t deliver commercial returns on such massive spending. Moving to a paid API creates a sustainable revenue channel for these hundreds of billions in AI capital expenditures, while the low-price strategy is designed to attract developers and build ecosystem scale.
Q4: Why did Meta’s stock only rise 4.7% after the AI launch?
Investor focus has shifted from "launching AI models" to "whether AI commercialization can convert into real revenue." The market’s doubts about Meta center on three areas: the certainty of API revenue contribution, the sustainability of $125 billion-level capital expenditures, and the timeline for AI investments to turn into profit.
