Meta’s latest results show two very different sides of the AI boom.
The company’s business continues to grow quickly. Second-quarter revenue increased 28% year over year to $60.8 billion, supported by the advertising engine behind Facebook and Instagram. At the same time, Meta’s free cash flow fell 91%, from $8.55 billion a year earlier to just $784 million.
Investors focused heavily on the second number. Meta shares dropped after the results as the company raised the lower end of its 2026 capital expenditure forecast to $130 billion, while keeping the upper end at $145 billion.
The reaction illustrates a shift in how the market evaluates AI. Building more infrastructure and training larger models are no longer enough on their own. Investors increasingly want to understand when that spending will produce returns.
Meta’s spending reflects the enormous infrastructure requirements behind modern AI.
Data centers, GPUs, networking equipment, energy, AI researchers, and model training all require significant investment. Meta needs that capacity not only for its AI models but also for AI features across Facebook, Instagram, WhatsApp, advertising, and future products.
CEO Mark Zuckerberg said Meta expects a significant portion of its computing capacity to support model training, its existing businesses, personal AI agents, and new products. The company also expects eventually to build a business serving large external customers.
However, many of those opportunities remain under development. The costs are immediate, while some of the potential revenue may arrive years later.
That gap is becoming increasingly important.
The difficult part of evaluating Meta’s AI investment is that AI is not producing value through a single product.
It is already deeply integrated into advertising and content recommendations. Better recommendation systems can increase engagement, while AI advertising tools can help businesses create campaigns, find audiences, and improve performance.
Meta’s 28% revenue growth suggests that its core business remains strong. However, separating the value created by AI from the value generated by the existing advertising platform is much harder.
This creates an interesting measurement problem. AI can improve an established business without immediately creating a completely new revenue stream.
For Meta, the question is whether those improvements can eventually justify infrastructure spending measured in hundreds of billions of dollars.
Retailers and brands obviously operate with very different budgets, but the underlying question is familiar.
AI tools are rapidly entering product search, content creation, customer service, merchandising, recommendations, logistics, and catalog management. It is relatively easy to launch a pilot or add an AI feature. Measuring whether it creates enough value is more difficult.
A product-content team, for example, might use AI to generate descriptions faster. The useful measurement is not simply how much content the model produced. Businesses also need to know whether the information remained accurate, whether manual correction decreased, and whether products reached sales channels faster.
Similarly, an AI shopping assistant should not be judged only by the number of conversations it handles. Retailers need to understand whether it improves discovery, conversion, customer satisfaction, or another meaningful outcome.
AI adoption, therefore, needs to move beyond capability to measurable performance.
There is another factor behind AI efficiency that receives less attention than computing power: the quality of the information being processed.
An e-commerce AI system working with incomplete or inconsistent product information has to infer missing details, reconcile conflicting attributes, or generate content without a reliable reference. That can create errors and require additional human review.
Structured product data gives these systems a stronger starting point.
Accurate specifications, identifiers, categories, descriptions, and multilingual content can support AI search, recommendations, enrichment, and customer service without requiring the model to reconstruct basic product facts every time.
For businesses evaluating the return on AI, better data can therefore be part of the efficiency equation.
Meta can afford a scale of experimentation unavailable to almost any retailer. Yet the market reaction to its results provides a useful lesson for the rest of the digital economy.
The question surrounding AI is gradually changing from “How much are you investing?” to “What are you getting from it?”
For e-commerce businesses, answering that question requires clear use cases, reliable data, and measurements connected to actual business outcomes. Faster content production matters if it improves time to market. Better recommendations matter if they help customers find the right products. Automation matters if it reduces repetitive work without introducing new errors.
Meta’s enormous AI bill puts the issue into sharp focus. As AI becomes a normal part of business infrastructure, increasing spending will not automatically create more value. The companies that benefit most will be those that can connect AI investment to measurable improvements in how their businesses actually operate.
Read further: News, AI, e-commerce, ecommerce, Icecat, Meta, product content
