For the better part of two years, Wall Street has watched Meta pour tens of billions of dollars into data centers and Nvidia chips with a single anxious question: where is the return? On Wednesday, the company floated an answer that sent its stock soaring. Meta is developing plans for a cloud business that would sell access to its excess AI computing capacity, and potentially hosted access to its models, to outside customers, according to reports from Bloomberg and others. The reaction was immediate and euphoric.
Meta shares jumped roughly 9% on July 1, adding an estimated $150 billion to the company's market capitalization in a single session. It was a striking reversal for a stock that had underperformed the S&P 500 in 2026, dragged down by investor unease over a capital spending program of historic scale.
The Bet Behind the Pop
The logic that excited investors is straightforward. Meta has told investors it expects to spend between $125 billion and $145 billion on capital expenditures in 2026, up from $72.2 billion in 2025, largely to build AI data centers and buy next-generation GPUs. That is nearly a doubling year over year, and more than the company spent in 2024 and 2025 combined. When Meta raised the top end of that guidance during its Q1 earnings call in April, the stock tumbled more than 6% in after-hours trading. The market wanted proof that the spending would pay off.
A cloud business offers exactly that narrative. Meta currently uses most of its infrastructure internally to train and run models like Llama and to power its advertising and recommendation systems. But it does not use all of it. According to Jefferies research cited by analyst Brent Thill, Meta's internal infrastructure utilization rate sits around 65%, leaving roughly 35% of idle capacity that could be rented out. Selling that surplus would raise utilization, improve return on invested capital, and generate cash flow from assets Meta has already bought.
Zuckerberg himself had teased the possibility. On the company's May earnings call, he said a Meta cloud computing business was "definitely on the table" as a way to generate returns on the AI buildout, while noting Meta expects to consume most of its capacity internally. The new effort is reportedly being steered by a heavyweight trio: head of infrastructure Santosh Janardhan, Meta Superintelligence Labs leader Daniel Gross, and president Dina Powell McCormick.
The SpaceX Playbook
Analysts quickly reached for a comparison that has become a template for the era: monetizing infrastructure built for a primary mission by renting out the spare capacity.
"This is very similar to the situation SpaceX has found itself in, which led it to sell compute capacity as well," said Gil Luria, managing director at D.A. Davidson. SpaceX, which built its Colossus supercomputers, has helped establish a market for hyperscale GPU capacity leased to major AI developers, with reported deals such as Anthropic leasing Colossus 1 capacity at $1.25 billion per month and Google agreeing to pay $920 million per month for Colossus 2 capacity. Those figures set a reference price for exactly the kind of asset Meta is sitting on.
Meta's entry pits it against the entrenched cloud giants: Amazon Web Services, Microsoft Azure, and Google Cloud, which together dominate a market worth hundreds of billions. But the sharpest pain landed elsewhere. Shares of neocloud specialists CoreWeave and Nebius cratered, falling roughly 13% to 15% each, as investors bet that Meta becoming a supplier could turn it from a customer into a competitor.
"The impact of adding Meta's capacity to the market is more likely to be on neoclouds than the big hyperscalers," Luria said. "Those companies like CoreWeave and Nebius rely on Meta for their growth and Meta may not need them anymore."
Why It Matters
Meta's move is the clearest signal yet of a broad shift in how Big Tech thinks about its AI infrastructure: not just as a cost of competing, but as a product. The compute-monetization trend reframes the trillion-dollar capex race. If a company can build far ahead of its own needs and then rent the overhang, the enormous fixed costs start to look less like a bet and more like a business.
That reframing is why investors cheered a mere report of exploratory plans. It offers a template for justifying spending that has otherwise strained margins and rattled shareholders across the sector. But the enthusiasm carries caveats. As CNBC noted, entering cloud means preparing for lower blended margins than Meta's advertising business commands, and running a cloud service is an operational discipline distinct from serving ads. The plans remain in development and the strategy could change.
There is also a bigger question lurking beneath the rally. If every hyperscaler with excess capacity becomes a seller, the scarcity that currently props up compute prices could erode. Jefferies' Thill argues demand still outstrips supply for now, but a wave of new sellers, from Meta to SpaceX to the neoclouds, could eventually test that assumption and, some worry, expose whether the AI infrastructure boom is overbuilt.
What to Watch
The immediate tell will be whether Meta formalizes the effort with pricing, timelines, and named customers, and whether management addresses it directly on the next earnings call. Watch, too, for how Meta's spending on third-party neoclouds evolves. If it pulls back on CoreWeave and Nebius contracts, the competitive threat becomes real rather than theoretical. And watch the margin math: investors rewarded the top-line promise, but a durable rerating will depend on whether Meta can turn idle GPUs into profit without dragging down the company that made it worth $150 billion more in an afternoon.
"This is very similar to the situation SpaceX has found itself in, which led it to sell compute capacity as well."— Gil Luria, Managing Director, D.A. Davidson