The Ad Engine Is Still the Cash Cow
When Meta reports its earnings, as it is expected to this week, Wall Street will be looking for impressive numbers. Analysts anticipate second-quarter revenue to land around $60.2 billion, a significant jump from the previous year. It’s easy to credit
this success to the company's highly publicized pivot to artificial intelligence. The reality, however, is that nearly all of this revenue—over 97%—comes from its deeply entrenched and massively profitable advertising business within the 'Family of Apps' like Facebook and Instagram. The story of Meta’s current financial strength isn’t about building futuristic AI agents; it’s about using existing AI to make its old business smarter, faster, and more efficient. The company’s ad machine is healthier than ever, not because it’s selling new AI products, but because it’s using machine learning to get better at its core function: selling our attention to advertisers.
The Two Kinds of AI at Meta
To understand the gap between perception and reality, you have to distinguish between the two types of AI at play. First, there's the profitable, workhorse AI. This is the machine learning that powers Meta's Advantage+ ad suite, optimizes ad placements in Reels, and predicts which users are most likely to click or buy. This AI is a cost-saver and a revenue-booster, helping the company recover from challenges like Apple's privacy changes by building sophisticated new ways to target users effectively without relying on old tracking methods. Then there’s the second kind of AI: the expensive, moonshot, generative AI. This is the quest for 'personal superintelligence' that Mark Zuckerberg talks about. This involves building massive, foundational AI models and the infrastructure to support them—a venture that is currently a colossal cost center. This part of the business is not yet profitable. In fact, it's swallowing capital at an astonishing rate.
The Billion-Dollar Burn Rate
While the ad business prints money, Meta’s other divisions are burning it. The Reality Labs division, home to its metaverse ambitions, consistently posts multi-billion-dollar losses each quarter on comparatively tiny revenue. Now, add to that the gargantuan spending on AI infrastructure. For 2026, Meta has projected capital expenditures between $125 billion and $145 billion, a figure that has investors spooked. These billions are being spent on data centers and specialized chips needed to train and run the next generation of AI models. It’s a long-term bet, and even CEO Mark Zuckerberg has admitted that the company’s ambitious AI agent development hasn't progressed as quickly as hoped and that the bets haven't come to fruition yet.
A Profitable Ad Company, Not a Profitable AI Company
So, what does this all mean? It means Meta is a phenomenally successful advertising company that is using AI tools to fortify its primary business. Its earnings are a testament to the enduring power of its social networks and its ability to refine its ad-targeting engine. However, this success should not be confused with proof that building large-scale, general-purpose AI is an inherently profitable business today. For now, it’s an incredibly expensive R&D project being funded by the immense profits from showing you ads on Instagram and Facebook. The road to monetizing generative AI directly—whether through AI agents, new creative tools, or even selling cloud compute power—is long and uncertain. Analysts and investors are right to be focused on capital expenditures, because that number tells the real story: the future is being built, but the present is paying for it.
