谷歌計劃1800億美元AI資本支出激增

• Google's infrastructure investments span data center buildout, subsea cable networks (Pacific, Africa, global routes), and fiber-optic connectivity through the America-India Connect initiative[2]
• Gemini AI models serve 750M+ monthly active users, requiring massive compute capacity across distributed data centers[1]
• Google Cloud's enterprise platform (Vertex AI) and Google Cloud services are primary beneficiaries of capex, supporting the $240B backlog[1]
• Infrastructure supports multiple AI use cases: government services (iGOT Karmayogi platform serving 20M+ public servants), scientific research partnerships via Google DeepMind, and climate technology initiatives[2]
• The capex surge reflects supply-constrained markets where hyperscalers report demand exceeds available compute capacity[3]

The $660-690B collective capex commitment from five major cloud providers signals an industry-wide conviction that AI workloads will consume all available compute capacity, potentially reshaping global infrastructure investment patterns. This spending surge ushers in a new era of AI infrastructure development comparable to historical utility and telecommunications buildouts[1][3]. However, the critical risk is whether AI revenue growth can justify this unprecedented investment; pure-play AI vendors (OpenAI, Anthropic) show rapid growth but their combined revenues remain a fraction of the infrastructure spending deployed on their behalf[3]. Success depends on sustained enterprise and consumer demand for AI services, with Google's strong cloud growth and Gemini adoption suggesting positive near-term momentum. The Stargate project ($500B by 2029) adds additional capacity, potentially intensifying competition and infrastructure redundancy[3].