Meta Taps Alexandr Wang for AI Development, Zuckerberg to Lead Sales

One year after a landmark $14 billion investment to integrate Alexandr Wang and a cadre of his top Scale AI engineers into its operations, Meta is undeniably back in the artificial intelligence arena. However, the tech giant still trails behind formidable competitors like OpenAI, Anthropic, and Google in this rapidly evolving market.

Wang’s most significant early achievement was the unveiling of the Muse Spark AI model in April. This release marked a pivotal shift for Meta, signaling a move away from its deep-rooted commitment to open-source (or more accurately, open-weight) models towards proprietary foundation models. The newly established Meta Superintelligence Labs, spearheaded by Wang, was specifically created to inject renewed dynamism into Meta’s AI endeavors, positioning it within one of the tech industry’s most intensely competitive sectors.

With CEO Mark Zuckerberg now armed with a proprietary model, the onus is on him to translate this technological leap into tangible financial success. This necessitates demonstrating Meta’s ability to attract paying users for its AI-driven tools, moving beyond simply leveraging AI to enhance its core advertising revenue streams.

“Meta needs to provide more compelling evidence of both adoption and commercialization,” stated Ralph Schackart, an analyst at William Blair who maintains a buy rating on the stock. “Investors are keenly looking for Meta to monetize new, AI-first products, beyond the substantial positive impact AI is already having on optimizing their advertising models.”

The market’s reaction thus far has been tepid. Meta’s stock has seen a 18% decline over the past twelve months, making it the laggard among megacap technology firms, alongside Microsoft, which faces its own set of AI-related hurdles. This performance is particularly noteworthy given Meta’s robust first-quarter revenue growth of 33%, the fastest expansion rate recorded since 2021.

Meta’s strategic misstep, in retrospect, began with its initial foray into AI. The company championed its Llama family of models with an open-source ethos, allowing developers extensive freedom to innovate. This approach stood in stark contrast to the monetization strategies employed by other major AI players, who typically charged for access.

The release of Llama 4 in April of the previous year failed to generate significant developer enthusiasm, prompting Zuckerberg to re-evaluate the company’s AI development trajectory. Two months later, in a move that surprised the tech world, Zuckerberg announced Meta’s $14.3 billion acquisition of nearly half of Scale AI, and crucially, brought Wang and his executive team aboard.

Wang’s instrumental role in the development and launch of Muse Spark in April of this year set the momentum in motion. “Instead of prioritizing third-party developers, the new model was engineered for seamless integration into Meta’s flagship applications like Facebook and Instagram, as well as its suite of AI-powered devices, such as the Ray-Ban Meta smart glasses,” explained Thomas Randall, an analyst at Info-Tech Research Group. This integration extends to the standalone Meta AI app and website.

“We’re seeing a dynamic shift among frontier model providers, and Meta requires a robust, reliable proprietary model that it wholly controls,” Randall elaborated. He added that Meta’s strategic pivot and substantial investments in Wang and other prominent AI hires over the past year, which Randall characterized as a “strategic rebuild,” were essential for the company’s future.

Randall conceded that Meta may not have taken the “most optimized route,” but he acknowledged, “I can now discern a clear vision for what they aim to achieve and what Wang is working towards.”

Following the debut of Muse Spark, Meta has introduced new AI-driven subscription plans, signaling an effort to diversify its business model beyond its traditional reliance on online advertising. Historically, this diversification has proven challenging, with Meta’s revenue still deriving approximately 98% from ads.

Schackart emphasized the need for “tangible evidence of a growing portfolio of new, AI-first products powered by Muse Spark, even if monetization takes time to materialize.” He believes this is precisely what investors are seeking.

Regardless of Muse Spark’s technical merits, Zuckerberg faces a significant uphill battle in regaining developer trust following the Llama misstep.

“I believe the AI community largely overlooks Meta at this juncture,” commented Rob May, CEO of Neurometric, a startup specializing in token engineering.

May noted the difficulty in assessing Wang’s leadership effectiveness within Meta Superintelligence Labs, citing the limited release of a single AI model that he described as underwhelming within the AI community due to its restricted accessibility.

While Meta actively courted third-party developers with Llama, May observes that the company’s current focus under Wang appears to be on internal applications. He recounted a period of regular communication with Meta regarding Llama-related matters, but now states, “I can’t get them to respond to messages.”

May conceded that Meta’s strategic emphasis on AI for its core advertising products is logical, given the substantial $200 billion annual business it seeks to protect. “They have built a formidable engine,” he stated.

Andrew Moore, CEO of enterprise startup Lovelace and former chief of AI at Google Cloud, believes it’s not too late for Meta to carve out a distinct niche.

Meta’s focus on enhancing model efficiency through advanced training techniques could serve as a significant differentiator for developers concerned about the escalating costs associated with foundation models. “If they can deliver proprietary, computationally efficient models, that would represent a stark contrast to the intense competition among the major players,” Moore suggested. “This could prove highly advantageous.”

Moore further indicated that Meta must demonstrate a clear advantage, whether in terms of cost, latency, or other technical nuances that are critical to developers.

Krish Subramanian, CEO of consulting firm KOI AI and a former product lead at IBM Consulting, observed that developers generally express more enthusiasm for Google’s AI models compared to Meta’s current offerings. He pointed out that Llama’s appeal lay in its direct targeting of developers seeking open-weight alternatives, a direction Meta has not significantly pursued with Muse Spark.

“The lack of developer trust will become a significant impediment if they don’t re-prioritize third-party developer engagement,” Subramanian cautioned, drawing a parallel to Microsoft’s lengthy journey to regain trust from open-source coders during the nascent stages of Azure. “Solely focusing on a walled-garden ecosystem and ad revenue as the primary income source will likely prevent them from becoming a dominant player,” he concluded.

A Meta spokesperson highlighted Wang’s recent remarks regarding the company’s ongoing commitment to the open-source ecosystem. They confirmed that Meta still intends to provide external developers with access to Muse Spark’s foundational technology via an API, as previously announced.

“We are currently engaged in testing with select early partners and anticipate a release this month,” the spokesperson stated.

Beyond the challenges with developers, Meta is also grappling with internal morale issues. The company has implemented significant workforce reductions throughout the year, including the dismissal of approximately 8,000 employees in May. These layoffs have impacted various departments, including teams responsible for trust and safety, raising concerns about potential implications for AI development, according to individuals familiar with the matter who requested anonymity to speak candidly.

Meta declined to provide comment on the layoffs. On the subject of safety, the spokesperson directed inquiries to Wang’s recent statements, where he emphasized, “One of the most critical aspects for me is the safety of these models,” during an appearance on the Core Memory podcast last month.

Tensions are also reportedly present at the leadership level of the AI division. While the Muse Spark release was met with internal acclaim, sources close to the situation indicate that Wang, along with former GitHub CEO Nat Friedman (who also joined last summer as part of the AI investment spree), face considerable pressure to deliver substantial revenue growth from both this model and future iterations.

Andrew Bosworth, Meta’s Chief Technology Officer and a 20-year company veteran, is a close confidant of Zuckerberg and could potentially assume a more prominent role in the company’s AI initiatives if the newer leadership is perceived as underperforming, according to these sources. Wang, however, has downplayed any reported internal conflicts, characterizing them as unfounded during his recent podcast appearance.

Wang has described Muse Spark as a preliminary offering, an “appetizer” for more powerful, “larger models” that are yet to come.

However, the AI community has grown accustomed to a consistent stream of updates and new features from established players like OpenAI, Anthropic, and Google.

“My primary concern is the frequency and cadence of these launches,” stated Howard Yu, a business professor at the International Institute for Management Development in Switzerland. “When a product is launched, can that momentum be sustained and built upon?”

Randall of Info-Tech Research Group believes that ultimately, the responsibility rests with Zuckerberg to define this strategy and demonstrate “how much of a superpower they have truly become with all of their products.”

Yu concurred, stating, “This is fundamentally a leadership challenge, isn’t it? Particularly in technology companies, the CEO is the one who articulates and defines the vision, especially when it involves multi-billion dollar investments.”

Yu further noted that Zuckerberg’s metaverse and virtual reality ambitions, which have resulted in over $80 billion in total losses since late 2020, complicate the AI narrative for investors. “He is running out of leeway for his credibility to remain intact,” Yu commented. “I believe the virtual reality endeavor may have significantly eroded his goodwill among investors.”