Meta spent $14.3 billion to bring in Alexandr Wang and his top engineers from Scale AI, betting that a proprietary AI strategy could revive the company's standing in artificial intelligence. A year later, Meta has delivered its first major model, Muse Spark, but investors and developers remain unconvinced that the company can compete with OpenAI, Anthropic, and Google in the race for AI dominance.

The stakes are high. Meta's stock has underperformed every other major tech company over the past 12 months, dropping 18 percent despite reporting 33 percent revenue growth in the first quarter of 2026, the fastest expansion rate since 2021. The disconnect reveals a fundamental investor concern: Wall Street doesn't believe Meta can build a profitable AI business beyond its core advertising model, which still accounts for 98 percent of the company's revenue.

What Went Wrong With Meta's Open-Source AI Strategy?

Meta's troubles in AI didn't start with Wang's arrival. The company initially pursued an open-source approach with its Llama family of models, allowing developers to freely use and modify the technology. This strategy backfired spectacularly. When Meta released Llama 4 in April 2025, the model failed to captivate the developer community, and the company's open-weight approach couldn't compete with paid alternatives from better-funded competitors.

That failure prompted Zuckerberg to make a dramatic strategic pivot. In June 2025, he announced the $14.3 billion investment to acquire roughly half of Scale AI and, more importantly, to bring Wang and his top lieutenants into Meta's newly created Superintelligence Labs. The move signaled that Meta was abandoning its bet on open-source AI and moving toward proprietary models designed for internal use and monetization.

Wang's first major deliverable was Muse Spark, released in April 2026. Unlike Llama, Muse Spark was designed to integrate directly into Meta's consumer products, including Facebook, Instagram, the Meta AI app, and AI-powered devices like Ray-Ban Meta glasses. The model received positive internal reviews, but external reception has been lukewarm at best.

Why Are Developers Ignoring Meta's New AI Model?

The core problem is trust. After the Llama debacle, the AI developer community has largely written off Meta as a serious player in frontier AI. Rob May, CEO of startup Neurometric, captured the sentiment bluntly: "I think the AI community largely ignores Meta at this point". He characterized Muse Spark as a "yawn" among AI professionals, partly because the technology isn't widely accessible to third-party developers.

Rob May, CEO of startup Neurometric

This represents a dramatic reversal. When Meta was promoting Llama, the company actively courted third-party developers. Under Wang's leadership, the strategy has shifted inward. May noted that he used to be in regular contact with Meta engineers about Llama-related issues, but now says he "can't get them to return messages".

May

The shift makes business sense in one respect. Meta has a $200 billion annual advertising business to protect, and focusing AI resources on enhancing ad targeting and content recommendation is a rational use of capital. But it comes at a cost: the company is ceding the developer ecosystem to competitors who are investing heavily in third-party relationships.

"The lack of developer trust will come back to hit them if they don't focus on third-party developers. To just focus on a walled-garden kind of an ecosystem and ad revenue as the main source of income, they probably will never become the big player," said Krish Subramanian, CEO of consulting firm KOI AI and former product head at IBM Consulting.

Krish Subramanian, CEO of KOI AI and former product head at IBM Consulting

Meta has promised to address this concern. A company spokesperson stated that Wang has publicly committed to continued support for the open-source ecosystem, and that Meta plans to release API access to Muse Spark's underlying technology for outside developers. The company said it was already testing with early partners and expected to release the API "this month".

What Would It Take for Meta to Win Back Investor Confidence?

Wall Street analysts have outlined specific benchmarks they want to see. Ralph Schackart, an analyst at William Blair who recommends buying Meta stock, said the company needs to "provide more proof points of both adoption and commercialization". He emphasized that investors are looking for Meta to monetize "a new AI-first product, beyond the substantial positive impact AI is having on enhancing the advertising models".

Meta has begun experimenting with new revenue streams. Since Muse Spark's release, the company has unveiled new AI and business-related subscription plans designed to diversify revenue beyond advertising. However, these efforts remain nascent, and there's no evidence yet that they're gaining meaningful traction with users.

Some analysts see potential differentiation opportunities if Meta executes correctly. Andrew Moore, CEO of enterprise startup Lovelace and former Google Cloud AI chief, noted that Meta has focused on making its models more computationally efficient through advanced training techniques. If the company can deliver proprietary models that cost significantly less to run than competitors' offerings, that could appeal to developers worried about rising foundation model expenses.

"If they do proprietary, computationally efficient models, that will be so different from what's happening in this death match between the big guys. They might really benefit," said Andrew Moore, CEO of Lovelace and former Google Cloud AI chief.

Andrew Moore, CEO of Lovelace and former Google Cloud AI chief

How to Understand Meta's AI Strategy Going Forward

• Proprietary Model Focus: Meta is shifting from open-source Llama models to proprietary Muse Spark, designed for integration into Meta's consumer products and devices rather than third-party developer use.
• Internal Monetization First: The company is prioritizing AI enhancements to its core advertising business, which generates 98 percent of revenue, while exploring new subscription-based AI products as secondary revenue streams.
• Developer Ecosystem Rebuild: Meta has committed to releasing API access to Muse Spark for third-party developers, though the company's track record with Llama has eroded developer trust significantly.
• Computational Efficiency: Meta is investing in training techniques that make its models more efficient and cheaper to run, potentially offering a cost advantage over competitors' larger, more expensive models.

The broader challenge for Zuckerberg is that hiring Wang and investing $14 billion was the easy part. Proving that Meta can build a sustainable, profitable AI business separate from advertising is far harder. The company faces skepticism from developers who remember Llama, investors who question the return on massive capex spending, and competitors with deeper expertise in frontier AI research.

Thomas Randall, an analyst at the Info-Tech Research Group, offered a more optimistic assessment, saying that Meta would have been "lost" without Zuckerberg's investment in Wang and other AI talent. He characterized the hiring spree as a necessary "strategic rebuild" for the company, even if the execution path hasn't been optimal.

For now, the outcome remains uncertain. Meta has the resources, the talent, and the user base to potentially succeed in AI. But success requires more than a good model; it requires winning back the trust of developers, convincing investors that AI can generate meaningful new revenue, and executing a strategy that balances internal product enhancement with external ecosystem participation. Zuckerberg's job, as the headline suggests, is to sell what Wang builds. So far, the market isn't buying.