Meta cloud plan raises questions over AI data center economics

Meta shares rose after a report that the company is preparing to sell spare AI computing power to other businesses, a move that would place it closer to the cloud market dominated by Amazon, Microsoft and Google. The reaction highlights investor appetite for AI infrastructure, even as some technology executives question whether Meta is suited to the business.

Fortune reported that Meta stock climbed more than 7% Tuesday after Bloomberg said the company is building a unit to offer excess AI compute capacity to outside customers. Bloomberg’s report described a potential new line of business that would put Meta up against Amazon Web Services, Microsoft Azure and Google Cloud.

Mark Douglas, president and CEO of connected-TV advertising platform MNTN, told Fortune that the bigger issue is the cost of building AI data centers in the United States. Douglas, who comments regularly on technology and media strategy, said U.S. data center capacity may look less appealing within two years because of high costs and local opposition.

Gulf capacity could pressure U.S. projects

Douglas said investors should pay more attention to Gulf-backed AI infrastructure, particularly data center capacity linked to Saudi Arabia’s Public Investment Fund. He argued that Saudi Arabia can use domestic energy to power data centers at lower cost than U.S. developers face.

According to Douglas, some Gulf data center projects are also being designed as legal extraterritorial zones that could help multinational clients meet data-residency requirements. He described those structures as data “embassies,” allowing companies to use overseas infrastructure while addressing rules tied to where data is considered to be held.

Douglas told Fortune he would avoid investing in companies building hyperscale capacity in the United States under current conditions. His view contrasts with the market response to the Bloomberg report, which rewarded Meta before any cloud-compute business had been tested at scale.

Questions over fit with Meta’s core business

Douglas also questioned whether selling compute capacity fits Meta’s existing business. He told Fortune the plan could be more about attracting attention than building a durable strategic line, and he compared it with efforts by SpaceX and xAI to rent out infrastructure to a small group of large buyers.

He said the customer base for raw data center capacity is limited because few companies can use it directly. If Meta tries to sell cloud services in the broader style of AWS, Azure or Google Cloud, Douglas said it would face an established and competitive market.

Douglas argued that shifting from a consumer social-media company with billions of app users to a vendor serving a small number of infrastructure customers would be an awkward fit. He also told Fortune that large companies often struggle when they try to enter new markets, citing Meta’s Threads as an example of the company pursuing adjacent opportunities.

AI still seen helping advertising

Douglas said his skepticism about a cloud-compute business does not extend to Meta’s broader AI work. He told Fortune he remains positive on Meta’s AI models and Llama, saying the company has capital, recruiting ability and large data resources that could help its advertising business.

MNTN is already talking with AI companies about advertising-focused models for targeting, Douglas said. He argued that stronger prediction tools can lower marketing costs and, in his view, pass savings back to consumers through lower product prices.

Douglas also rejected the idea that Meta needs a cloud business to become less dependent on advertising. He told Fortune that Meta is already the world’s largest social media company and said people expect it to become the world’s largest advertising company this year, making an AI infrastructure push more of a tactical detour than a central strategy if it happens.

This story draws on original reporting from Fortune.