Meta Platforms (META -2.43%) is once again releasing some of its artificial intelligence (AI) models with open weights, allowing developers to download and modify them.
The company released Muse Glimmer, a 30-billion-parameter model, on Aug. 10. However, as of Aug. 25, the company had not yet released the weights for its more powerful flagship model, Muse Spark 1.2.
So, does Meta Platforms' heavy investment in AI infrastructure make sense if it plans to make more of its AI models publicly available?
Image source: Getty Images.
Earning money from AI without selling models
Unlike OpenAI and Anthropic, Meta Platforms does not need to rely primarily on subscriptions, application programming interfaces (APIs), and enterprise software to monetize its AI models directly. Instead, the company is leveraging AI to improve recommendations and advertising across Facebook and Instagram. The company is also developing business messaging and AI-agent opportunities through WhatsApp and Messenger.
NASDAQ: META
Key Data Points
Meta Platforms' AI strategy is already showing results. In the second quarter (ended June 30), improvements to its AI models generated an 8.3% increase in Facebook ad clicks and a 15.7% rise in ad conversions. Additionally, an early Instagram AI pilot increased the number of users completing targeted in-app actions by 1%.
Heavy investment
Meta Platforms' revenue grew 28% year over year to $60.8 billion, while operating cash flow was $31.9 billion in the second quarter. With the company's capital expenditures reaching $31.1 billion, free cash flow was only $784 million. Meta Platforms now expects $130 billion to $145 billion in capital spending in 2026, up from its initial $115 billion to $135 billion range.
Meta Platforms' spending commitments also extend well beyond 2026. The company had $81.6 billion of noncancelable contractual commitments due in 2027, mostly associated with cloud capacity and investments in technical infrastructure, data centers, and Reality Labs hardware. The company also entered into another $68 billion in data center leases, expected to begin in 2027 and 2028.
Meta Platforms' high spending could also weigh on future profit margins. Depreciation expense associated with property and equipment rose 40% year over year to $6 billion in the second quarter. Construction in progress also reached $80.3 billion, with most of those costs related to data centers, servers, and network infrastructure. As those assets come online, depreciation expenses could rise further, putting pressure on operating margins.
Reasonable valuation
Meta Platforms is currently trading at around 16.8 times analysts' expected 2027 earnings per share of $33.9 (as of Aug. 25). The valuation appears reasonable, considering the company may not need significant user growth for its AI investments to pay off. With 3.6 billion people already using Meta Platforms' apps daily, even relatively small improvements in user engagement and ad conversion can potentially translate into meaningful revenue gains.
However, this advantage has a limit. Much of the growth in the number of ads Meta shows is coming from lower-monetization markets such as Asia-Pacific. Hence, local-language business agents are particularly important, since they may help the company generate more revenue from users in markets where advertising currently monetizes at lower rates.
So, AI needs to continue lifting revenue per user while capital spending and depreciation eventually moderate. Meta Platforms' forward earnings multiple leaves room for upside. But if AI spending remains elevated without a corresponding improvement in advertising monetization, the stock may struggle to grow.
