NVIDIA's AI Infrastructure Dominance: Growth Metrics, Cloud Economics, and Market Outlook
NVIDIA's data center revenue reached $26B with compute revenue up 5x and networking up 3x, capturing approximately 33% of major cloud providers' capex budgets. The company's profitability has surged dramatically, with net income reaching $29.8B in fiscal 2024 and a 62% LTM margin. GPU unit economics show $1 spent on NVIDIA infrastructure generates $5-7 in cloud provider revenue over 4 years, with demand for H200 and Blackwell chips exceeding supply.
Metrics in this report
38%
benchmark
Amazon Web Services operating margin
40b
total
Major cloud providers' data center capex spending
400-450k
range
Cost per DGX H100 unit
26b
total
NVIDIA fiscal 2024
5revenue ratio
current
Cloud provider revenue over 4 years per $1 NVIDIA infrastructure spend
7revenue ratio
projected
Cloud provider revenue over 4 years per $1 H200 spend
2400users
simultaneous
NVIDIA HGX H200 server capacity
24000tokens per second
single server
NVIDIA HGX H200 server performance
40%
current
Proportion of NVIDIA data center revenue from inference workloads (trailing 4 quarters)
42.6b
trailing 12 months
NVIDIA latest trailing 12 months
62%
trailing 12 months
NVIDIA latest trailing 12 months
13b
total
NVIDIA data center revenue from major clouds (45% of total)
350000H100 GPUs
by end of 2024
Meta's planned total GPU acquisition
24000H100 GPUs
total
GPU count used to train Meta's Llama 3 model
33%
estimated
Of major cloud customers' total capex budgets
29.8b
total
NVIDIA fiscal year 2024
10xmultiple
dollar terms over 2 years
NVIDIA profit growth from 2022 to 2024
57%
total
NVIDIA fiscal year 2024
35000GPUs
total
NVIDIA GPUs deployed in Tesla training cluster
60%
current
Proportion of AI workloads dedicated to model training