Meta's Massive Artificial Intelligence Investments May Be About to Pay Off for Investors
Meta Platforms has invested heavily in artificial intelligence, establishing the Meta Superintelligence Lab and acquiring Alexandr Wang for $14 billion. The new Muse Spark model, while competitive, is not yet superior to rivals. However, it could significantly enhance Meta's advertising business, leading to increased ad impressions and revenue. The Muse model is more cost-efficient, potentially reducing compute costs by 90%. This could allow Meta to scale AI features effectively, attracting more advertisers and boosting revenue, making its current stock valuation appear favorable.
Meta Platforms (META +1.60%) established its new artificial intelligence (AI) lab last year, the Meta Superintelligence Lab, spending billions of dollars on personnel and compute capacity to rebuild its artificial intelligence models from the ground up. That includes the $14 billion acquisition of Alexandr Wang, who heads up the lab. Nearly a year later, it finally has something to show for it.
Meta's new Muse Spark model is the first new model from Wang's team. It's competitive with leading models from OpenAI, Anthropic, and Alphabet's (GOOG 0.12%) (GOOGL 0.03%) Google on most benchmarks, but it's far from blowing the competition away. In fact, it lags competing models in most regards.
But Meta could still see a huge boost in profits thanks to its new model.
Image source: Getty Images.
The importance of artificial intelligence to Meta
Meta is perhaps the company best positioned to benefit from continued advances in artificial intelligence. Advances in AI can improve Meta's advertising business, its content recommendations, its ability to attract new advertisers, its users' ability to create new and interesting content, its messaging business, and more.
The positive impact of Meta's AI efforts is already substantial, as seen in its 2025 earnings results. Meta's ad impressions across its family of apps increased 12% last year. At the same time, the price per ad increased 9%. Those two numbers historically move in opposite directions. The rising engagement on its platform and the ease of optimizing ad campaigns have led to improvements in both numbers so far.
If Meta wants to continue improving those numbers, though, it needs to use better models. Licensing a model from a leading AI lab, however, would be cost-prohibitive at Meta's scale. It's far more efficient in the long run for it to build out its own AI models tailored specifically for its needs.
That's certainly capital intensive; Meta spent $72 billion on capital expenditures last year with plans to spend as much as $135 billion this year. But the release of its new Muse model could mean those investments are about to pay off.
A step change in Meta's AI
One of the most important details from Meta's Muse Spark release is how much it costs to run compared to Meta's latest Llama model. "We can reach the same capabilities with over an order of magnitude less compute than our previous model, Llama 4 Maverick," the company said in a press release.
That means Meta could spend just 10% of what it currently spends on compute for generative AI without any degradation in its capabilities. In practical terms, Meta will likely spend more than that, but see improved capabilities that lead to higher engagement and ad targeting, resulting in a further increase in ad revenue.
Additionally, Muse Spark may enable it to more efficiently scale AI features like business AIs and chatbots it's testing in Mexico and the Philippines with great success. That could lead to further increases in advertising through click-to-message ads and additional monetization potential for WhatsApp and Messenger.
NASDAQ: META
Key Data Points
Muse Spark could help Meta deliver on its goal of offering a fully fledged advertising agent that can develop and manage ad campaigns for small businesses with AI-generated creatives, copy, and targeting. That could attract more advertisers, further boosting ad revenue.
It's worth looking at Alphabet's Google Search results to see the potential of AI improvements on large existing advertising businesses. Alphabet produced incremental improvements in its AI models throughout 2025, ultimately releasing Gemini 3.0 in November. It increased the rollout of AI Overviews as the cost of providing AI-generated responses decreased with improved model efficiency.
Google also saw a bump in engagement as users found its search engine could answer more questions, and it saw improved ad targeting as its model could better understand search intent. The result was an acceleration in ad revenue throughout the year, including a 17% increase in the fourth quarter.
Meta could experience similar results as introducing new AI features and scaling them become less cost-prohibitive with its new models. An acceleration in revenue could offset the spike in operating costs it's seen from its AI hiring spree while leading to better-than-expected earnings. That makes the stock's current valuation of just 21.5 times analysts' earnings expectations look like a bargain.
