Goldman Sachs: Broadcom Reiterates AI Revenue Target of 230 Billion, 1GW Cluster Generates 30 Billion Annually

According to Chaotiang Research, a Goldman Sachs research report dated September 8, 2026 notes that Broadcom reaffirmed its FY27/FY28 AI revenue targets of $115 billion and $230 billion at the CommunaCopia conference, stating that supply, not demand, remains the primary constraint. Land acquisition, power capacity, and facility construction for data centers represent the true bottlenecks to AI deployment, whereas supply chain constraints remain relatively manageable. Broadcom’s $35 billion XPV partnership with Blackstone and Apollo can support over 20 gigawatts of AI compute capacity while capping Broadcom’s direct financial exposure. A 1-gigawatt AI cluster generates approximately $30 billion in annual revenue, with annual operating costs of around $10 billion. Broadcom believes that the economic value of AI primarily flows to the owners of frontier models. Annual industry training expenditure stands at approximately $200 billion. Frontier models capture roughly 75% of industry revenue ($120 billion in revenue against $100 billion in costs), while open-source models generate only about $30 billion in revenue despite similar investment levels. Goldman Sachs maintains a Buy rating, setting a target price of $540 based on 30x normalized earnings per share of $18, implying approximately 51% upside from the current stock price.