Meta Eyes Cloud Business to Challenge AWS, Azure, Google

Meta’s cloud computing business went from rumor to a stated corporate option on Wednesday, when chief executive Mark Zuckerberg told shareholders that renting out the company’s enormous computing power is definitely on the table. The remark, delivered at Meta’s virtual annual shareholder meeting on May 27, is the clearest signal yet that the owner of Facebook, Instagram and WhatsApp could one day compete with Amazon Web Services (AWS), Microsoft Azure and Google Cloud.

What Zuckerberg described, though, was narrower than the headline suggests. He pitched a cloud business as a fallback for the day Meta decides it has built more data center capacity than its own apps and artificial intelligence models can absorb, a hedge against a record spending bill of $125 billion to $145 billion this year rather than a product with a launch date.

Zuckerberg Put a Cloud Business ‘on the Table’

A shareholder asked the obvious question: would Meta take on the established cloud providers? Zuckerberg did not commit to anything. He did confirm that the demand already exists. Companies approach Meta most weeks, he said, asking it to stand up an API (application programming interface, the connection layer that lets outside software tap a provider’s models and servers) or to buy raw processing capacity outright.

Almost every week there are different companies that come to us from outside asking us to both stand up an API service or asking if we have compute that they could buy from us at some premium to what we’ve bought it at. We haven’t done that yet because we think that we have a use for the compute. Obviously if we get to a point where we feel that we have overbuilt, then that is an option that we have.

Those words matter because Meta is the only one of America’s big infrastructure spenders that has never rented its servers to anyone else. AWS, Azure and Google Cloud were each built, in part, to sell spare capacity to outsiders. Meta’s machines have spent two decades doing one job: powering its own consumer apps and, more recently, its open Llama model family. You can hear the full exchange through Meta’s 2026 annual shareholder meeting webcast.

The Capex Curve Driving the Pitch

The cloud comment did not arrive in a vacuum. It landed weeks after Meta lifted its 2026 capital expenditure (capex, the money a company sinks into long-lived assets like buildings and chips) guidance, citing higher component prices and additional data center costs. That spending plan is close to double what Meta laid out the year before, and it has unsettled some investors who want to see the return.

The anxiety is not hypothetical. When the company first flagged the higher range in late April, the stock slid roughly 7% in a single session. Zuckerberg’s pitch to those same shareholders was straightforward: if the buildout overshoots demand, the surplus compute becomes a product Meta can lease, which turns a feared liability into a revenue line.

• $56 billion in first-quarter 2026 revenue, up about 33% from a year earlier, the cash engine funding the buildout.
• $125 billion to $145 billion earmarked for 2026 capital spending, raised from a prior range of $115 billion to $135 billion.
• $600 billion the figure Zuckerberg has floated for AI infrastructure over the coming years, dwarfing any single year’s budget.

The detail of the guidance sits in Meta’s first-quarter 2026 results filing with the SEC, which ties the higher number to component pricing and capacity for future years.

The Compute Meta Already Owns

If Meta does flip the switch, it would not be starting from a standing start on hardware. The company is assembling some of the largest computing sites on the planet, anchored by two multi-gigawatt campuses that are already under construction or close to opening.

Two Megasites Under Construction

Project Hyperion in Louisiana and Project Prometheus in Ohio are the headline assets, and they differ sharply in scale and timing. Hyperion is the long-haul bet; Prometheus is the one arriving first.

Silicon and Custom Chips

Behind the buildings sits the silicon. Meta has spoken about amassing compute equivalent to well over a million high-end GPUs (graphics processing units, the chips that train and run AI models), most of them bought from NVIDIA. It has also built its own line of custom accelerators, the Meta Training and Inference Accelerator (MTIA), to handle parts of its workload more cheaply. That is the raw material a rental business would sell. It is also, for now, fully committed to running Meta’s own services.

What Meta Still Lacks to Sell Cloud

Owning the chips is the easy part. Selling cloud to banks, hospitals and government agencies is a different company, and this is where the gap between Zuckerberg’s option and a real product is widest. A hyperscale rental business runs on machinery Meta has never had to build.

• No enterprise sales force tuned to multi-year corporate contracts, procurement cycles and the support obligations that come with them.
• No mature compliance stack: the single-tenant isolation, data-sovereignty options and security certifications that regulated buyers demand before they trust a vendor with sensitive workloads.
• A live conflict with Google Cloud, with whom Meta holds a large infrastructure arrangement; standing up a rival service would strain a relationship it currently depends on.

None of these is impossible to acquire. AWS spent years building its enterprise muscle before it dominated. But each takes time and money, which is why a cloud launch reads less like a near-term threat and more like a switch Meta is keeping within reach. The compliance hurdle alone has kept newer entrants out of finance and public-sector deals for years.

AWS, Azure and Google Cloud Hold the Ground

Even a fully equipped Meta would be walking into a market where the three incumbents are entrenched and, in two cases, growing fast. In the first quarter of 2026, global cloud infrastructure spending rose about 35% from a year earlier to roughly $129 billion for the period, according to Synergy Research Group, and the big three took more than two-thirds of it.

The growth column is the part that should worry a new entrant. Azure and Google Cloud are not coasting; they are expanding faster than the market and pouring their own billions into AI capacity. A breakdown of the global cloud infrastructure market share over time shows how durable the leaders’ positions have been. Meta would be arriving late to a contest the incumbents have spent fifteen years learning to win.

The Subscriptions Arriving Before the Servers

While the cloud option stays on the shelf, Meta is moving on a faster way to turn its AI spending into income: charging consumers directly. The company said it will begin testing paid AI tiers under its Meta One brand in a handful of pilot markets next month, starting with Singapore, Guatemala and Bolivia.

• Meta One Plus ($7.99 a month): priority access to high-fidelity image generation, video creation and longer-context reasoning tools.
• Meta One Premium ($19.99 a month): advanced multimodal features plus priority processing when traffic spikes.

The logic is the same one Zuckerberg gave the cloud question. Pairing a software-as-a-service (SaaS, subscription software delivered over the internet) consumer fee with the long-term option to lease the back-end gives Meta two routes to recoup its outlay rather than one. The subscription play is the bird in the hand; the cloud business is the one still in the bush. You can track how the strategy develops through Meta’s investor relations annual-meeting hub.

If Meta’s own apps and models keep swallowing every server it builds, the cloud business stays a line in a shareholder answer, and Azure customers have nothing new to weigh. If the buildout outruns that internal demand, the same surplus that scares investors becomes the inventory of a fourth hyperscaler, and the option Zuckerberg sketched on Wednesday turns into a sales pitch.

Frequently Asked Questions

Is Meta launching a cloud service to compete with Azure and AWS?

Not yet. Zuckerberg called a public cloud business an option that is on the table if Meta finds it has overbuilt its AI data centers. There is no product, price list or launch date, and Meta has never rented its servers to outside customers before.

When could Meta enter the cloud market?

No timeline has been given. Zuckerberg framed it as conditional on Meta having spare capacity beyond what its own apps and AI models need. Until the company decides it has overbuilt, the cloud business stays a contingency rather than a plan.

How much is Meta spending on AI infrastructure in 2026?

Meta has guided to capital expenditure of $125 billion to $145 billion for 2026, raised from an earlier range of $115 billion to $135 billion. Zuckerberg has separately floated a figure of around $600 billion for AI infrastructure over the next several years.

What is Meta One?

Meta One is the company’s subscription brand spanning Facebook, Instagram and WhatsApp. Its new paid AI tiers, Meta One Plus at $7.99 a month and Meta One Premium at $19.99 a month, are being tested first in Singapore, Guatemala and Bolivia.

Why would competing in cloud be hard for Meta?

Meta owns the chips and the data centers but lacks an enterprise sales force, the security and data-sovereignty certifications regulated buyers require, and the freedom to compete head-on with Google Cloud, an existing infrastructure partner. Building those would take years.