Meta released its first enterprise AI agent in June 2026 to help businesses with customer service and daily operations [1]. CEO Mark Zuckerberg said Meta sees a large opportunity selling AI services such as APIs, business agents, and potentially compute capacity directly to enterprise customers [1, 2].
Unlike major cloud providers, Meta does not currently operate a cloud infrastructure business to sell compute power [3, 2]. However, the company is exploring launching a business to lease AI compute to third parties, with ongoing talks involving startups like Anthropic [3]. Zuckerberg acknowledged Meta has received offers to sell AI compute at significant premiums but stressed it would be "foolish to basically just sell all of the compute and take a short-term profit" [3, 2]. Meta aims to balance monetizing compute today with preserving capacity for future AI development [1, 3, 2].
In Q2 2026, Meta sharply raised its capital expenditure guidance to $130 billion-$145 billion for the full year, driven largely by expanding AI data center infrastructure [3]. The heavy investment caused free cash flow to plunge 90-91% year-over-year for the quarter [3, 2]. Zuckerberg highlighted that much of Meta’s core business is already benefiting from AI, delivering more relevant apps and improved business outcomes [4].
Since early 2026, Meta has launched multiple AI-assisted standalone apps including Forum (Groups), Seller (Marketplace), and Instagram Instants as well as experimenting with AI-generated bedtime stories [4]. Its Threads app, powered by AI-driven recommendations, reached 500 million monthly active users [4].
Meta’s next step includes continuing to expand enterprise AI offerings while carefully managing its AI compute resources amid growing demand and internal use cases [1, 3, 2].
