Meta’s free cash flow dropped by 91% in the second quarter, as investments in AI infrastructure entered a phase requiring significant capital expenditure.

The competition in the field of generative AI is shifting from the announcement of new models to considerations related to financial statements. Meta’s latest quarterly figures show that, even as its core advertising business continues to grow, the massive investment required for building data centers, purchasing chips, and developing advanced models significantly reduces the amount of cash available for the company’s use.

According to a Reuters report dated July 30, Meta’s free cash flow in the second quarter was $784 million, a decrease of around 91% compared with $8.55 billion in the same period last year. The company has adjusted its forecast for capital expenditures in 2026 to between $130 billion and $145 billion, up from the previous range of $125 billion to $145 billion. Meta expects to invest up to $145 billion this year in AI infrastructure, which is roughly twice the amount spent in the previous year.

Meanwhile, Meta’s core business operations remained stable. Revenue in the second quarter increased by 28% on a year-on-year basis, reaching 60.8 billion dollars, while the number of daily active users of its apps amounted to 3.6 billion, a 3% increase compared to the previous year. This indicates that the issue is not whether investments in AI generate sufficient revenue, but rather whether the speed at which new businesses generate cash flows can keep up with the pace of expansion of the infrastructure.

During the earnings call, Zuckerberg emphasized that computing power will be used to train models, improve core business operations, develop personal AI assistants, and serve large customers. Meta aims to turn these personal AI assistants into a new consumer-oriented business, leveraging its large user base and advertising distribution capabilities. However, until the business model is fully matured, costs related to depreciation, electricity, networking, and maintenance will continue to appear in the financial statements.

This change has also redefined the advantages that large technology companies possess. In the past, internet platforms could expand rapidly by relying on software with low capital requirements; but in the era of AI, leaders need to have capabilities in model development, chip supply, data center construction, and energy acquisition. Capital efficiency, the utilization rate of data centers, and the cost of model inference have become key indicators for assessing the quality of an AI strategy.

For the market, Meta’s quarterly performance does not indicate that its investments in AI have been unsuccessful; rather, it serves as a reminder to the industry that there is a costly transition period between achieving leadership in model development and generating sustainable profits on a large scale. In the coming quarters, it will be up to personal agents to improve ad conversion rates, generate subscription revenue, or open up new opportunities in the enterprise services sector – these factors will determine the return on these investments.

Source:Reuters