After spending $37 billion building out AI data centers last year, Facebook and Instagram maker Meta Platforms expects to spend around $70 billion in 2025, with more in the coming years. Even with all that capital spending, it still will need more computing power, this week signing a $14 billion cloud contract with CoreWeave that goes through 2031.
Unlike Amazon.com, Microsoft, and Alphabet, which rent out their AI servers in the cloud, Meta’s buildout is all for its own use. There are no sales to cloud customers to offset its capital expenditures.
Meta’s balance sheet and cash-flow statement are shifting as a result. In the second quarter, Meta had more debt and lease liabilities than cash and short-term investments for the first time as a public company. Free cash flow—operational cash flow minus capex—was down 16% from 2024 in the first half of the year.
Meta is hardly in any real financial risk, but all the AI spending is starting to bite, and investors want to know when they will see a return. CEO Mark Zuckerberg can read the room as well as anyone, and in the first-quarter earnings call in April, he devoted the bulk of his scripted presentation to laying out what Meta stands to gain from all its investment.
He talked about Meta AI, its ChatGPT competitor, and how it will follow Meta’s usual path to monetization. “First, we build and scale the product, and then once it is at scale, then we focus on revenue,” he said. “But I expect that we’re going to be largely focused on scaling and deepening engagement for at least the next year before we’ll really be ready to start building out the business here.” Zuckerberg returned to this theme in the second-quarter earnings call in July.
For Meta, revenue has always meant ads. Most recently we’ve seen this path with Meta’s X competitor, Threads, which launched in July 2023. By January 2025, Threads had over 300 million monthly users, and Meta began inserting ads.
Now it has that kind of scale for Meta AI, which has at least one billion monthly users. This week Meta announced that it would begin to use their AI chats to better target ads in the company’s social media apps. Ads in AI chats themselves may not be far off.
All of this is delicate, given the increasingly personal data users have begun to share with chatbots.
It’s possible that ChatGPT maker OpenAI actually beats Meta in crossing the chatbot-advertising Rubicon. OpenAI is still investor funded, and it has committed to future obligations that add up to hundreds of billions of dollars. It needs to show how it intends to eventually become profitable.
While Meta can lean on substantial cash flows from Facebook and Instagram, OpenAI’s subscription and user-based revenue are growing quickly but they can’t cover ChatGPT’s mounting expenses, especially since the vast majority of ChatGPT’s 700 million weekly users continue to use a free version of the app.
In May, OpenAI hired Fidji Simo for a newly created role, CEO of Applications, reporting to CEO Sam Altman. Simo rose through the ranks at Meta to eventually run the Facebook app. She then became CEO of Instacart, overseeing the delivery firm’s initial public offering. She still chairs Instacart’s board and sits on the boards of Shopify and OpenAI.
Simo is an executive with deep understanding of advertising and e-commerce. The company declined to answer questions about her new role or to make Simo available for comment, but she is the exact kind of person a company would hire as it pivots to a free service backed by advertising and e-commerce.
We saw OpenAI’s first steps in that monetization journey this past week. In partnership with payment processor Stripe, ChatGPT users can now look for and buy items from Etsy, without ever leaving the chatbot, through a shopping experience ChatGPT calls Instant Checkout.
Millions of Shopify stores will be next to join ChatGPT’s e-commerce platform. Once e-commerce shows up, advertising isn’t far behind, with sellers looking to get attention.
All of this will feel familiar to anyone that lived through the rise of Web 2.0 sites like Google and Facebook. At first, investors funded the search and social-media sites, allowing the free services to grow by leaps and bounds with users. Once the monetization switch gets turned on, the companies refocused their efforts on advertisers and sellers. Eventually, users and advertisers are all hooked on the platform, and the companies turn their attention to pleasing shareholders.
Expect a similar pattern for AI, with one key difference: The shareholder payoff could take time, with initial profits wiped out by hundreds of billions in spending.
Write to Adam Levine at adam.levine@barrons.com