Meta Plan to Rent Out AI Infrastructure Fuels Fears Chip Demand Has Peaked
Summary
- Meta said it will lease out entire data center computing capacity through its infrastructure as a service (IaaS) business.
- Meta’s plan to rent out large-scale server capacity fueled fears that semiconductor demand could weaken, as other AI companies may find it more economical to rent infrastructure than buy new semiconductor chips.
- With Meta spending $145 billion in CAPEX to build AI infrastructure, expectations for a new cash pipeline helped send Meta’s shares up 8.8%.
Forecast Trend Report by Period


Leasing Out Entire Data Center Computing Capacity
Renting Beats Buying Chips
Market Sees AI Investment Race Nearing a Peak

Meta Platforms said on July 1 it plans to expand into two cloud businesses: platform as a service, or PaaS, and infrastructure as a service, or IaaS.
The market reaction focused on IaaS. Under that model, Meta would lease out the raw computing power of data centers equipped with Nvidia’s top-end graphics processing units, or GPUs, to external customers. The service is aimed at big tech companies and large AI startups that want to advance their own large AI models but lack enough physical servers and chips. The move brings Meta into an infrastructure rental market that has been led by so-called neocloud companies such as CoreWeave and Nebius. CoreWeave shares sank 14.0%, while Nebius fell 12.3%, reflecting investor fears.
Through its PaaS business, Meta plans to combine AI models with infrastructure. It would offer access to its in-house AI model, Muse Spark, running in advance on Meta-built high-performance infrastructure. Outside developers and companies could apply Meta’s AI technology to their services through application programming interfaces, or APIs, without having to build complex data centers and servers themselves. That would set up a direct challenge to established products such as Amazon Web Services’ Bedrock and Microsoft Azure’s AI Foundry.
The two business lines could directly affect semiconductor demand. If Meta begins offering large-scale server capacity to the market, other AI companies may find it more economical to rent Meta’s infrastructure than to buy expensive new chips. That has fueled fears the semiconductor market could shift from supply shortage to oversupply. Meta, which had already signaled $145 billion in capital spending this year for AI infrastructure, is now saying it has excess capacity to put on the market. That has prompted speculation the infrastructure spending race among big tech companies has hit its peak.
Meta’s shift appears to benchmark a model used earlier by Elon Musk’s xAI. xAI recently signed long-term deals to lease computing resources from Colossus, its massive data center in Memphis, Tennessee, to rivals including Anthropic and Google. Industry participants estimate xAI spent about $30 billion building Colossus, but could generate nearly $30 billion in annual revenue from leasing alone, allowing it to recoup that investment in just one year.
That helps explain why Meta shares rose even as SOXX, a leading exchange-traded fund tracking the Philadelphia Semiconductor Index, fell 4.7%. Expectations for a new cash pipeline drove Meta shares up 8.8% on the day.
Kim In-yeop in Silicon Valley and Park Shin-young in New York, correspondents for Korea Economic Daily, inside@hankyung.com
Korea Economic Daily
hankyung@bloomingbit.ioThe Korea Economic Daily Global is a digital media where latest news on Korean companies, industries, and financial markets.