A market research firm has released a report rebutting concerns over Meta's excessive investment, which arose from recent news that the Facebook parent plans to sell some of its artificial intelligence (AI) infrastructure to outside parties. The report projected that Meta would build its own AI platform even as it enters the cloud business.
Market research firm SemiAnalysis said in a report released Tuesday that "even after entering the cloud business, Meta's investment in data centers and AI computing will not slow down but rather accelerate further." According to the report, excluding the ultra-large data centers Meta is building on its own, the company secured more than 5 gigawatts (GW) of power capacity through cloud and data centers in the first half of this year alone.
SemiAnalysis paid particular attention to Meta's push to collaborate with AI startup Anthropic and its AI model "Claude." This suggests the firm sees a possibility that Meta could secure the foundation to service Claude on its own infrastructure. SemiAnalysis also raised the possibility that Meta could, over the long term, develop into a platform offering not only Anthropic but also other AI models such as OpenAI. "Meta possesses powerful distribution networks including massive graphics processing units (GPUs), data centers, and social platforms," SemiAnalysis said. "It could leap forward as a key player in the AI platform and application distribution market, rather than a mere GPU-buying company."
Earlier, Bloomberg reported Monday that Meta had internally launched a "Meta Compute" plan and was devising a cloud business model utilizing its data center infrastructure. According to sources, Meta is pursuing the cloud business in two broad ways. First, it would place its latest in-house AI model "Muse Spark" on its own infrastructure and allow external developers to use it through an application programming interface (API). This is a platform-as-a-service (PaaS), similar to Amazon Web Services' (AWS) "Bedrock," Microsoft (MS) Azure's "AI Foundry," and Google Cloud's "Vertex AI."
Another method is to lease out, in bulk, the surplus resources among the computing infrastructure built for developing artificial superintelligence (ASI), such as data centers. This is a business type similar to the infrastructure-as-a-service (IaaS) model mainly used by companies such as CoreWeave and Nebius. If Meta enters the cloud business, a seismic shift is expected in a market that has been divided among Amazon Web Services, Microsoft Azure, and Google Cloud. Meta is the only one among the four major hyperscalers—large-scale AI data center operators including Amazon, Microsoft, and Google—that does not provide cloud services.
However, the news of Meta's entry into the cloud business was also read as a signal that "there are surplus resources in AI infrastructure." As a result, a considerable number of semiconductor-related stocks underwent sharp price corrections on Sunday and Monday.
