Hua Er Jie Jian Wen·2025-09-30 13:46Core Insights - Citigroup's latest research indicates that investment and deployment in AI infrastructure are accelerating at a pace far exceeding expectations, driven by genuine enterprise-level demand, distinguishing it from the 2000 internet bubble [1][4][14] - The forecast for AI capital expenditure has been significantly raised, with projections for 2026 increased from $420 billion to $490 billion, and cumulative capital expenditure by 2029 raised from $2.3 trillion to $2.8 trillion [1][4] - The expected growth rate for capital expenditure in 2026 is 24%, notably higher than the current market consensus of 20% [4] Investment Trends - Recent collaborations among industry giants highlight the rapid scale and speed of AI infrastructure development, with significant deals such as Oracle's $300 billion agreement with OpenAI [5][7] - The acceleration in infrastructure investment is fundamentally driven by end-user demand, as reflected in the strong order backlog data from major tech companies [8][14] AI Infrastructure Demand - AI infrastructure demand is experiencing exponential growth, with frontier model training computing power increasing at a rate of 4.6 times per year since 2010, surpassing Moore's Law [6] - Despite improvements in GPU cluster performance, hardware costs for leading AI supercomputers are rising exponentially, with costs increasing by 1.9 times annually and training power requirements by 2.1 times [6] Future Projections - By 2030, global demand for AI computing power is expected to add 55 GW of electricity capacity, translating to an additional $2.8 trillion in AI computing expenditure, with the U.S. market accounting for $1.4 trillion [8][14] Distinction from Internet Bubble - The current AI investment wave is fundamentally different from the internet bubble, as there is a clear external demand driven by enterprises adopting AI services, providing a reliable validation of value [14] - Unlike the past, where many internet companies spent heavily on marketing to attract non-existent customers, AI companies today have visible and reliable demand levels in various sectors such as knowledge retrieval, customer service, and healthcare [14]