拆解OpenAI的AI需求后，巴克莱得出结论：AI资本开支周期仍将持续，技术突破可能在27/28年引发算力需求激增

Core Insights - OpenAI's revenue performance significantly exceeds internal expectations, indicating rapid growth in AI demand and a sustained capital expenditure cycle in the short term [1][2] - The report suggests that the risk of a bubble burst in the AI sector remains low as long as OpenAI maintains its strong growth trajectory [1] Revenue Performance - OpenAI's actual revenue for 2025 is projected to be approximately 15% higher than mid-year forecasts, while the 2027 revenue expectation has been raised by 50% [2] - The total revenue forecast for OpenAI in 2027 has been adjusted from $60 billion to $90 billion, with computing costs increasing from $21 billion to $30 billion [2] - Weekly active users (WAU) are expected to rise from 1.4 billion to 1.8 billion, and the average revenue per paid user (ARPU) is anticipated to increase from $748 to $880 [2] - CEO Sam Altman has indicated that OpenAI aims to achieve $100 billion in annual recurring revenue (ARR) by 2027, a year earlier than previously predicted [2] AI Capital Expenditure Cycle - The sustained performance of OpenAI confirms that the AI capital expenditure cycle will continue in the medium to long term [3] - OpenAI's revenue is directly tied to computing power, with significant investments planned for computing operations, exceeding $450 billion from 2024 to 2030, peaking at around $110 billion in 2028 [3][6] - Continuous model iterations, such as GPT-6 and Sora 3, are driving up computing demands, necessitating accelerated infrastructure deployment by computing partners [6] Strategic Partnerships and Contracts - OpenAI has signed approximately $650 billion in computing power leasing contracts with various partners over the next decade [7] - Major contracts include $300 billion with Oracle OCI, $250 billion with Microsoft Azure, and additional agreements with Google GCP, Amazon AWS, and CoreWeave [7] Industry Competition and Long-term Strategy - The competitive landscape is intensifying, leading to an "arms race" among tech giants like Google and Meta to expand user bases and accelerate model iterations [7] - The global AI data center capacity is expected to double from 114.3 GW to 236 GW between 2024 and 2030, with OpenAI requiring partners to invest over $600 billion in computing clusters [7] - Tech giants are committed to long-term investments in AI, with founders emphasizing the importance of maintaining high capital expenditures despite market fluctuations [8]