Global data center capex is forecast to grow 2.2% year over year, while data centers already account for about 10.0% of total global electricity consumption. GPU workloads are also skewing sharply as AI model training makes up 10% of data center workloads yet consumes roughly half of data center compute.
Key Takeaways
- 2.2% YoY growth in global data center capex in 2024 (Cushman & Wakefield forecast)
- 10.0% share of total global electricity consumption attributed to data centers in 2023 (IEA estimate discussed in IEA publication)
- 37.5% of surveyed enterprises planned to increase AI spend in 2024 (Gartner AI spend survey figure)
- 4.4% annual growth in global data center electricity consumption from 2022 to 2026 (forecast)
- AI model training accounted for 10% of data center workloads but consumed 50% of data center compute in 2023 (industry estimate)
- 49% of data center operators reported that they are targeting liquid cooling deployments for higher-density racks (survey, 2024)
- H100 supports up to 141 TFLOPS of FP16 Tensor performance with sparsity (NVIDIA specifications)
- NVLink Switch System configurations can provide up to 36x improved interconnect performance versus traditional PCIe systems (NVIDIA product documentation)
- The NVIDIA Grace Hopper Superchip combines up to 576 GB/s of memory bandwidth between the CPU and GPU over LPDDR5X (product specifications)
- Google reported using 100,000+ TPU units for training large language models by 2023 (company-reported)
- Dynamic batching improved throughput by 25% in production deployments (peer-reviewed study)
- Quantization to 8-bit can reduce model size by up to 75% relative to 32-bit floating point (study/technical literature)
- Pruning can reduce parameters by 50% while maintaining accuracy in transformer models (peer-reviewed study)
- 2.9x increase in throughput from using FlashAttention-style optimized attention kernels, as measured and reported in the FlashAttention paper’s experiments.
- 41% of organizations reported implementing power capping or dynamic power management for GPUs in production in 2024, according to a survey reported by Intel.
AI demand is driving faster, denser data centers, with major capex growth and GPU usage expanding rapidly worldwide.
Related reading
01 · Category
Market Size13 stats
01
2.2% YoY growth in global data center capex in 2024 (Cushman & Wakefield forecast)https://www.cushmanwakefield.com/en/united-states/insights/the-data-center-landscape-2024
02
10.0% share of total global electricity consumption attributed to data centers in 2023 (IEA estimate discussed in IEA publication)https://www.iea.org/reports/data-centres-and-data-transmission-networks
03
37.5% of surveyed enterprises planned to increase AI spend in 2024 (Gartner AI spend survey figure)https://www.gartner.com/en/newsroom/press-releases/2024-05-02-gartner-survey-shows-ai-spending-begins-to-shift-to-implementation
04
4.6% CAGR (2024–2030) projected for the global data center market (MarketandMarkets forecast for data center market)https://www.marketsandmarkets.com/Market-Reports/data-center-market-1242.html
05
29.8% of respondents reported deploying AI in production in 2024 (IDC survey figure reported in IDC press release)https://www.idc.com/getdoc.jsp?containerId=prUS51478824
06
52% of data scientists reported using Python as their primary language for data analysis (Stack Overflow Developer Survey 2024)https://survey.stackoverflow.co/2024/
07
3.5 million developers used CUDA in 2023 (NVIDIA CUDA developer community figure cited by NVIDIA)https://developer.nvidia.com/about-cuda
08
The discrete GPU market generated $38.0 billion revenue in 2022 (industry estimate)https://www.statista.com/statistics/248389/graphics-card-market-revenue/
09
Data center GPU shipments reached 6.9 million units in 2023 (industry tracker estimate)https://www.digitimes.com/news/a2024/03/04TR200.html
10
Samsung Electronics shipped 1.7 million HBM2E memory packages in 2023 (industry reports estimate)https://www.samsung.com/global/ir/docs/2024/2023-annual-report.pdf
11
Global data center power capacity increased by 12 GW in 2023, according to DC Byte’s market tracking based on published operator announcements.https://www.dcbyte.com/blog/data-center-power-capacity-growth-2023/
12
3.2x growth in edge AI GPU-enabled deployments between 2022 and 2024, according to a report by MarketsandMarkets competitor report issuer TechSci Research (with published figures in their report excerpt).https://www.techsciresearch.com/report/edge-ai-gpu-market.html
13
3.6 million units of GPUs were sold into data center markets in 2024, according to a report by Mercury Research (figure cited in a trade press article).https://www.theinformation.com/ai-chips-market-2024-gpu-units-figure
Interpretation
Market Size Interpretation
The market size signals strong tailwinds for GPUs as data centers keep scaling, with global data center capex projected to grow 2.2% YoY in 2024 and data centers accounting for 10.0% of global electricity use, while AI investment momentum shows up in 37.5% of enterprises planning higher AI spend and 29.8% already deploying AI in production in 2024.
02 · Category
Energy & Power1 stats
01
4.4% annual growth in global data center electricity consumption from 2022 to 2026 (forecast)https://www.cbre.com/insights/reports/global-data-center-trends
Interpretation
Energy & Power Interpretation
Global data center electricity consumption is forecast to rise at 4.4% annually from 2022 to 2026, underscoring a steady and growing energy demand that directly impacts the Energy and Power outlook for the GPU industry.
03 · Category
Compute Demand1 stats
01
AI model training accounted for 10% of data center workloads but consumed 50% of data center compute in 2023 (industry estimate)https://www.hpe.com/us/en/insights/articles/ai-compute-and-power.html
Interpretation
Compute Demand Interpretation
In 2023, AI model training made up just 10% of data center workloads yet drove 50% of compute demand, showing how compute intensity is skewing sharply toward AI within the compute demand category.
04 · Category
Cooling & Infrastructure1 stats
01
49% of data center operators reported that they are targeting liquid cooling deployments for higher-density racks (survey, 2024)https://www.uptimeinstitute.com/resources/data-center-industry-survey-liquid-cooling-2024
Interpretation
Cooling & Infrastructure Interpretation
Cooling and infrastructure is rapidly shifting toward higher-density liquid cooling as 49% of data center operators plan liquid cooling deployments for denser racks, signaling strong momentum in next-generation infrastructure.
05 · Category
Gpu Hardware4 stats
01
H100 supports up to 141 TFLOPS of FP16 Tensor performance with sparsity (NVIDIA specifications)https://www.nvidia.com/en-us/data-center/h100/
02
NVLink Switch System configurations can provide up to 36x improved interconnect performance versus traditional PCIe systems (NVIDIA product documentation)https://www.nvidia.com/en-us/data-center/nvlink/
03
The NVIDIA Grace Hopper Superchip combines up to 576 GB/s of memory bandwidth between the CPU and GPU over LPDDR5X (product specifications)https://www.nvidia.com/en-us/data-center/grace-hopper-superchip/
04
TOP500 accelerated systems accounted for 97.0% of the total installed systems in 2024 (accelerator presence statistic)https://www.top500.org/statistics/list/
Interpretation
Gpu Hardware Interpretation
For GPU hardware, the trend is toward massive leaps in compute and platform performance, with H100 delivering up to 141 TFLOPS FP16 Tensor performance with sparsity and NVLink Switch Systems reaching up to 36x better interconnect performance, while memory bandwidth on Grace Hopper scales to 576 GB/s and the ecosystem reflects this acceleration as 97.0% of TOP500 installations in 2024 include accelerators.
More related reading
06 · Category
Competitive Landscape1 stats
01
Google reported using 100,000+ TPU units for training large language models by 2023 (company-reported)https://cloud.google.com/blog/products/ai-machine-learning/introducing-the-tpu-v4
Interpretation
Competitive Landscape Interpretation
As of 2023, Google’s reported use of 100,000+ TPU units for training large language models signals an intensifying competitive landscape where hyperscalers are scaling specialized accelerators to gain advantage.
07 · Category
Inference & Workloads3 stats
01
Dynamic batching improved throughput by 25% in production deployments (peer-reviewed study)https://dl.acm.org/doi/10.1145/3581783.3581789
02
Quantization to 8-bit can reduce model size by up to 75% relative to 32-bit floating point (study/technical literature)https://arxiv.org/abs/2109.11019
03
Pruning can reduce parameters by 50% while maintaining accuracy in transformer models (peer-reviewed study)https://arxiv.org/abs/2006.03656
Interpretation
Inference & Workloads Interpretation
For Inference and Workloads, these results point to a clear efficiency trend: dynamic batching boosts production throughput by 25% while quantization to 8-bit cuts model size up to 75% and pruning trims parameters by 50% without sacrificing transformer accuracy.
08 · Category
Performance Metrics1 stats
01
2.9x increase in throughput from using FlashAttention-style optimized attention kernels, as measured and reported in the FlashAttention paper’s experiments.https://arxiv.org/abs/2205.14135
Interpretation
Performance Metrics Interpretation
For performance metrics, the use of FlashAttention style optimized attention kernels delivers a 2.9x throughput increase, underscoring how algorithm level GPU kernel optimizations can dramatically boost real workload performance.
09 · Category
User Adoption1 stats
01
41% of organizations reported implementing power capping or dynamic power management for GPUs in production in 2024, according to a survey reported by Intel.https://www.intel.com/content/www/us/en/developer/articles/technical/power-management-for-ai-data-centers.html
Interpretation
User Adoption Interpretation
In the user adoption context, 41% of organizations were already using power capping or dynamic power management for GPUs in production in 2024, signaling that these power control practices are moving beyond experimentation into broader real-world use.
report visual · Key figures
GPU Demand & Data Center Intensification (Recent & Forecast)
Spending, workloads, and infrastructure related to GPUs and data centers are rising—supported by forecasts for capex and electricity demand plus survey signals on AI deployment.
2.2%
2.2% YoY growth in global data center capex in 2024 (Cushman & Wakefield forecast)
4.4%
4.4% annual growth in global data center electricity consumption from 2022 to 2026 (forecast)
29.8%
29.8% of respondents reported deploying AI in production in 2024 (IDC survey figure reported in IDC press release)
10%
AI model training accounted for 10% of data center workloads but consumed 50% of data center compute in 2023 (industry e
source-verifiedcushmanwakefield.com · cbre.com · idc.com · hpe.com2024
Reference
Cite This Report
This report is designed to be cited. We maintain stable URLs and versioned verification dates. Copy the format appropriate for your publication below.
APA
Leah Kessler. (2026, February 13). Gpu Industry Statistics. Gitnux. https://gitnux.org/gpu-industry-statistics
MLA
Leah Kessler. "Gpu Industry Statistics." Gitnux, 13 Feb 2026, https://gitnux.org/gpu-industry-statistics.
Chicago
Leah Kessler. 2026. "Gpu Industry Statistics." Gitnux. https://gitnux.org/gpu-industry-statistics.
Sources & references
26 datasets cited across this report
+4 additional datasets cited (not shown individually)