メタの新AIモデル「Muse Spark」は初期段階で有望だが、投資家はザッカーバーグ氏の戦略を重視する

watch now

With Mark Zuckerberg counting on Meta’s new artificial intelligence model, Muse Spark, to revive his company’s standing in the booming AI market, guidance and commentary are going to be of heightened importance following first-quarter earnings on Wednesday.

That’s because the new model, formerlycodenamed Avocado, was unveiled in early April, just as the second quarter was getting underway. Muse Spark marks a turning point in Meta’s AI strategy, moving it away from the previous Llama models, which were released for free to the open-source community.

Meta indicated that it eventually wants to make money from the technology by offering paid access to developers, similar to the approach being pursued by OpenAI, Anthropic and Google. What’s important today, analysts say, is that Meta’s AI tools continue to bolster its dominant ad business, and that the company shows its AI technology can compete with the market leaders.

According to Arena.AI, a site that tracks quality and performance of the top models, Meta AI trails Anthropic’s Claude and Google’s Gemini in text, but only Claude in vision, as of Sunday. In both areas it’s currently ahead of OpenAI’s GPT. Claude also leads in the categories of document and code, where Meta is further down the leaderboard.

In a report to clients last week, analysts at Citizens described AI as a “complementary good” for Meta, and said they expect to hear much more on the company’s earnings call.

“We are impressed with Meta’s Muse Spark model,” the analysts, who recommend buying the stock, wrote in the report, citing the model’s strength in text and vision. “While the company integrated Meta AI into its core apps, we are awaiting a strategy to drive scaled consumer usage that is akin to other AI chatbots like ChatGPT and Claude as we believe this can unlock new data and ad budgets.”

watch now

Meta’s ad business continues to grow, boosted by increased targeting capabilities that come with AI advancements. Analysts expect year-over-year revenue growth of 31% for the first quarter to $55.6 billion, according to LSEG. That would represent the fastest rate of expansion since 2021.

But Wall Street has been looking for momentum in AI beyond just advertising, as OpenAI and Anthropic have seen their combined valuations swell past $1 trillion thanks to the popularity of their AI models and services. Meta’s stock price is up 24% in the past year, while Alphabet shares have gained 116% over that stretch, boosted by the growth in Gemini.

When Meta revealed Muse Spark earlier this month, it was pitched as the first major AI model to be spawned from Meta Superintelligence Labs, led by Alexandr Wang, the company’s chief AI officer. Wang was previously CEO of Scale AI, and he joined Meta in June as part of the company’s$14.3 billion investment into the data-labeling startup.

Zuckerberg followed that up with more high-profile hires. He brought in former GitHub CEO Nat Friedman along with business partner Daniel Gross, who was previously the CEO of AI startup Safe Superintelligence, which Ilya Sutskever co-founded in 2024 after leaving OpenAI.

“This leadership shift and the subsequent nine-month rebuild of Meta’s AI stack signal an aggressive effort to close the gap with competitors like OpenAI (private) and Google,” Truist analysts wrote in a report on April 21. “Notably, Muse Spark is closed-source, reflecting a change from Llama’s open-sourced approach and a shift toward high-performance, specialized infrastructure.”

‘Back into the AI conversation’

Meta showed that its internal testing,released in conjunction with Muse Spark’s debut, indicated the model is less powerful than bleeding-edge AI models from Anthropic and others, a way for the company to manage early expectations.

Still, analysts have expressed relief that Meta is finally out of the gate, with more models presumably on the way. JPMorgan Chase analysts wrote in a report last week that Muse Spark “has brought Meta back into the AI conversation.”

“Investor sentiment on Meta is turning increasingly constructive,” the analysts wrote. “The stock has been pressured by elevated expenses and capex, concerns around AI model delays, and an adverse social media legal decisions.”

watch now

Meanwhile, Meta is cutting head count as it zeroes in on AI.

The company said Thursday that it wouldlay off 10% of its workforce, about 8,000 employees, on May 20, in an effort to improve business efficiencies. That’s happening as Meta pours money into AI infrastructure, telling investors in January that 2026 AI-related capital expenditures should come in the range of $115 billion to $135 billion, up from $72.2 billion in 2025.

Analysts at Loop Capital wrote in a recent report that Meta’s hefty investments have fed a negative perception that it’s “a company desperately spending to fix problematic AI initiatives.” The release of Muse Spark, they said, shows that Meta is producing AI models that could further improve its coreonline ad business.

Even if Muse Spark and future models from Meta fail to outperform rival systems, those tests are of “mixed importance,” because of the company’s clear advantage in ads, the Loop analysts wrote.

“Foundational LLM/agentic reasoning models are certainly key for Meta, but we view image/video generation models as strategically important with greater near-term engagement and monetization implications,” they wrote. “The real bar for success is building models that power excellent products for users, creators and advertisers.”

WATCH: Meta and Broadcom deepen AI chip partnership through 2029.

watch now