After investing more than $14 billion to recruit Alexandr Wang and a team of top engineers from Scale AI for an ambitious overhaul of its artificial intelligence initiatives, Meta has regained some visibility in the AI landscape. However, it continues to lag significantly behind industry leaders such as OpenAI, Anthropic, and Google. Wang's major achievement thus far is the launch of the Muse Spark AI model in April, which represents Meta's initial foray into proprietary foundational models, marking a shift away from an exclusive focus on open-source technology—often referred to as "open weight" in the AI field.
Wang heads the Meta Superintelligence Labs, a division created to inject excitement and vitality into the company’s AI presence at a time when the sector is rapidly evolving. With the rollout of Muse Spark, the responsibility now rests on CEO Mark Zuckerberg to turn this innovation into a profitable venture. This entails demonstrating that Meta can not only attract paying customers for its AI solutions but also leverage technology to bolster its core advertising operations. Ralph Schackart, an analyst at William Blair, stresses that for investors to regain confidence, they need to see substantial adoption and commercial success from Meta’s AI endeavors—beyond simply enhancing advertising models.
Despite reporting a remarkable 33% revenue growth in the first quarter—the fastest increase for any quarter since 2021—Meta’s stock has dropped 18% over the past year, making it one of the poorest performers among major tech companies, alongside Microsoft, which also faces challenges in the AI arena. The company’s initial approach to AI through the Llama series of models is viewed by some experts as a misstep. This open-source strategy allowed developers extensive flexibility, while competitors opted for paid access.
Meta’s introduction of Llama 4 in April last year failed to garner interest from developers, prompting Zuckerberg to rethink the company’s AI strategy. In a surprising move two months later, he announced a $14.3 billion investment in Scale AI, effectively bringing Wang and his core team onboard. Muse Spark has since created opportunities for integration into Meta’s applications such as Facebook and Instagram, as well as AI devices like the Ray-Ban Meta glasses.
Thomas Randall of the Info-Tech Research Group notes that the focus of Muse Spark is shifting towards Meta’s own platforms rather than external developers, a strategy that could risk leaving the company behind if it doesn't establish a stable and proprietary model to rival those of its competitors. He also remarked that the investments Zuckerberg made in AI talent were crucial for Meta’s future, allowing for a clearer vision of its objectives.
As part of its shift away from a reliance on advertising revenue—which accounts for 98% of its income—Meta has introduced new AI and business-related subscription services since the launch of Muse Spark. Schackart emphasizes the need for clear evidence of new AI-first products emerging from this latest model, even if monetization takes time.
Concerns linger in the developer community, especially following the issues surrounding the Llama models. Rob May, CEO of Neurometric, pointed out that Meta is generally overlooked by AI professionals and that Wang's leadership is still unproven, given the limited release of new models. While Meta has historically aimed to engage third-party developers, the current focus appears more internally driven, leading to waned interest from external collaborators.
Despite the hurdles, industry leaders like Andrew Moore, a former Google Cloud AI head, believe there’s still potential for Meta to carve out a niche. He argues that if Meta can develop proprietary and computationally efficient models, it could distinguish itself in a pool of giants competing fiercely. Yet, experts caution that developers are currently more excited about offerings from companies like Google than from Meta.
Meta has emphasized its ongoing support for the open-source community, indicating plans to provide access to Muse Spark's technology via an API, with initial testing already underway. However, a broader challenge remains—low morale amid steady job cuts, including a significant layoff of about 8,000 employees in May, spanning roles pivotal for trust and safety in AI.
Amidst ongoing changes within the AI division, pressure mounts on Wang and fellow leaders, such as former GitHub CEO Nat Friedman, to generate substantial revenue from Muse Spark and upcoming releases. Although high internal praise has followed the Muse Spark launch, the overarching strategy to ensure success ultimately lies with Zuckerberg, who has significant influence over the direction of the AI offerings.
With Wang describing Muse Spark as just a precursor to stronger models on the horizon, many in the AI community are eager for a more consistent rollout of innovations. Howard Yu, a professor at the International Institute for Management Development, emphasizes that building upon momentum from product launches is crucial, placing the onus on Zuckerberg to create a solidified vision and execution strategy for Meta’s positioning in the competitive AI landscape. The challenges posed by past ventures into virtual reality and the significant financial losses incurred therein may complicate the narrative as Meta seeks to regain investor trust.
