The most expensive infrastructure buildout in corporate history just found a possible second act. On Wednesday, CNBC’s Julia Boorstin reported that “Sources close to the situation do confirm that META is working on building a cloud infrastructure business to sell AI compute.” The stock responded, jumping 7.56% on July 1 as investors digested the idea that Meta Platforms (NASDAQ:META | META Price Prediction) might soon compete with the three companies that have owned enterprise AI compute for a decade.
Why Meta suddenly wants to be a cloud vendor
Start with the capex line. Meta guided full-year 2026 capital expenditures to $125 to $145 billion, raised from a prior range and blamed on “higher component pricing and additional data center costs.” Q1 capex alone was $18.997 billion, up nearly half year over year, all documented in Meta’s Q1 8-K filing. That is a lot of GPUs for a company whose revenue still overwhelmingly comes from selling ads against Reels and Stories.
Zuckerberg has been hinting at the release valve for a while. Per reporting around the plan, he has said “There are different companies that come to us from outside asking us… if we have compute that they could buy from us at some premium to what we bought it at.” And the strategic logic, in his own framing, is a hedge on overbuild. “If we get to a point where we feel that we have overbuilt, then that is an option that we have,” he said. Selling excess capacity turns a potential capex disaster into a business line.
What Meta is walking into
The incumbents are moving fast. Amazon (NASDAQ:AMZN) reported AWS revenue of $37.587 billion in Q1 2026, growing 28% year over year, its fastest pace in 15 quarters, with landmark commitments from OpenAI, Anthropic, and, awkwardly, Meta itself. Andy Jassy told investors “Our chips business topped a $20 billion revenue run rate” on Graviton, Trainium, and Nitro.
Microsoft (NASDAQ:MSFT) is arguably in the strongest position. Fiscal Q3 saw Azure and other cloud services grow 40%, an AI business at a $37 billion run rate up 123% year over year, and a commercial remaining performance obligation of $627 billion. Those are contracted future dollars.
Then there is Alphabet (NASDAQ:GOOGL), which grew Google Cloud revenue 63% to $20.028 billion with a backlog north of $460 billion, nearly doubling quarter on quarter. Google has also reportedly limited Meta’s access to its Gemini AI models due to computing constraints, a data point that makes the “build our own paid API service” ambition look necessary.
What Meta actually has to sell
Meta owns custom MTIA silicon, one of the industry’s largest fleets of NVIDIA GPUs, and Louisiana’s Hyperion campus. It also owns Llama, the most-downloaded open-weight model family, and just launched its first model from Meta Superintelligence Labs. A paid inference API where developers rent time on Meta’s chips against Meta’s models is a natural product. So is renting raw capacity to hyperscaler customers priced out elsewhere.
The optionality is real, but so is the credibility gap. Reality Labs still lost $4.03 billion in Q1 2026, and Meta shares are down 6% year to date and off 15% over the past year as investors price in the capex intensity. Enterprise cloud is a services business with SLAs, sales engineers, and procurement cycles Meta has never run.
The Q2 earnings call is the obvious venue for Zuckerberg to give this an actual name, pricing model, and go-to-market timeline. Until then, a cloud business exists mostly as a hedge against a capex number that keeps expanding each quarter.
Contact [email protected] for any questions or corrections.
Top Gaining Stocks
Top Losing Stocks