AI Infrastructure CapEx Surge Indicates Hyperscaler AI Compute Acceleration

Key findings

1. Over the past 48 hours, data points reveal a significant increase in AI infrastructure capital expenditure across hyperscalers, with notable investments in data centers, GPUs, and networking hardware, driven by AI compute and inference needs.
2. Microsoft, Meta, Google, and Amazon have reported record or elevated CapEx figures for Q4 2023 and FY 2024, emphasizing AI infrastructure acceleration, GPU cluster deployment, and network expansion as primary drivers.
3. Microsoft's AI infrastructure CapEx reached approximately $14 billion in Q4 2023, an 80% YoY increase, reflecting record quarterly spending linked to Azure and OpenAI integration. Meta's CapEx guidance for FY 2024 is between $30–37 billion, with a focus on GPU clusters for AI acceleration. Google’s CapEx is projected at a $32 billion run rate for 2024, with a notable step-function increase in data center and networking investments for Gemini 1.5. Amazon’s Q4 2023 CapEx of $16.8 billion is primarily allocated to AI training and inference infrastructure. Nvidia’s data center revenue proxy of $14.5 billion in Q4 FY24 shows a 409% YoY jump, indicating hyperscaler CapEx converting into GPU orders. Broadcom reports a backlog of $11 billion in AI networking orders for 2024, suggesting network-switch CapEx acceleration tied to AI clusters. Additionally, Dell and Equinix announced a new 20 MW AI-ready colocation expansion, reflecting enterprise spillover demand. The global data center CapEx forecast for 2024 is $400 billion, with 70% attributed to AI-driven builds, representing a 23% YoY increase.