Data-centre server: Use case
Use cases
Each use, its share of demand and volume, how critical the product is to it and why, and the price paid
| Use case | Share of demand | Volume | Unit | Criticality | Reason | Price | Price unit | Year | Source |
|---|---|---|---|---|---|---|---|---|---|
| Cloud, web and enterprise computing in hyperscale and colocation data centres | 74% | – | – | Essential | Cloud services run on servers, and firms cannot rent computing without a provider's servers somewhere. | – | – | 2023 | LBNL estimates that 74% of United States servers were in hyperscale and colocation data centres in 2023 [11] |
| Servers in firms' and governments' own computer rooms and small data centres | 26% | – | – | Important | Much of this work could move to rented cloud capacity, at a cost in control, latency or data rules. | – | – | 2023 | Calculated as 100% minus the 74% in hyperscale and colocation centres [11] |
| AI training and inference (GPU servers; overlaps the rows above) | 17% | 24.7 | billion USD of AI-optimised servers sold by Dell in its fiscal 2026; share is of 2026 forecast server shipments (units) | Essential | Large AI models need many GPUs working together, which general-purpose servers cannot supply at useful speed. | 270,000 | USD per DGX H100, 2023 | 2026 | [14][19][32] |
as of 2026-11-16
Industries that buy it
The industries and kinds of firm that buy it, for which use case, and how critical it is to them
Used by- Hyperscale cloud providers Cloud, web and AI services · criticality essential With colocation operators, they hold 74% of United States servers (2023) [11]; no source splits the two.
- Colocation and wholesale data-centre operators Servers placed by tenants in rented space · criticality essential Included in the 74% above [11].
- Foundation model developers Training and running AI models · criticality essential GPU servers are the main hardware for training; the share of AI servers they buy was not found.
as of 2026-11-16
What it competes with
The products and technologies that compete with it in each use case
Competes with- Cld Cloud computing Firms' and governments' own computer rooms Renting capacity from a provider replaces owned servers; the provider then buys the servers instead, so demand moves rather than disappears (74% of United States servers were in hyperscale and colocation sites in 2023) [11]. No cost comparison was found.
as of 2026-11-16
How it is used
Lifetime, use per year and utilisation, by variant
| Measure | Value | Unit | Variant | Place | Year | Source |
|---|---|---|---|---|---|---|
| Product lifetime assumed in the life cycle assessment of a Dell PowerEdge R760 | 4 | years | General-purpose | World | 2025 | [1] |
| Server lifetime assumed in LBNL's model of the United States installed base, 2023 (4.4 years for 2000 to 2019) | 5.7 | years | General-purpose | United States of America | 2023 | LBNL modelling assumption for United States servers [11] |
| Years of use in a hyperscale data centre before a downstream user, as assumed in ITRenew's model | 3 | years | General-purpose | World | 2020 | [28] |
| Expected economic useful life of compute servers in ITRenew's model, including a downstream user | 9.2 | years | General-purpose | World | 2020 | [28] |
| Typical energy use of a Dell PowerEdge R740 (typical energy consumption) | 1760.3 | kWh per year | General-purpose | World | 2019 | [2] |
| Operational power of a dual-processor server set in LBNL's model for 2023 | 600 | W | General-purpose | United States of America | 2023 | LBNL model value [11] |
| Measured average power of an 8-H100 node during training (78% of rated power) | 7.9 | kW | AI accelerator | World | 2024 | [11] |
as of 2026-11-16
Prices
Unit prices by market or use, with year and source
| Market | Value | Unit | Year | Source |
|---|---|---|---|---|
| NVIDIA DGX H100 (eight GPUs), selling price | 270,000 | USD per system | 2023 | SemiAnalysis estimate [32] |
| NVIDIA DGX H100 cost of components, before vendor margin | 45,000 | USD per system | 2023 | Calculated sum of four cost lines, 26.6 + 8 + 4 + 6.4 thousand USD [33] |
| NVIDIA DGX-1 (eight Tesla P100 GPUs) at launch | 129,000 | USD per system | 2016 | [31] |
as of 2026-11-16
Sources
- Life Cycle Assessment of Dell Server PowerEdge R760 (report produced September 2025) Dell Technologies, 2025
- PowerEdge R740 product carbon footprint (report produced January 2019) Dell Technologies, 2019
- 2024 United States Data Center Energy Usage Report (LBNL-2001637) Lawrence Berkeley National Laboratory, 2024
- Strong Demand from CSPs and Sovereign Cloud to Drive Over 20% Growth in AI Server Shipments by 2026, Says TrendForce (30 October 2025) TrendForce, 2025
- Dell Technologies Delivers Fourth Quarter and Full-Year Fiscal 2026 Results Dell Technologies via Business Wire, 2026
- Data Center Impact Report: The Financial & Sustainability Case for Circularity (Fenn and Fesch; the document shows no date) ITRenew, hosted by the Open Compute Project
- Nvidia announces a supercomputer aimed at deep learning and AI TechCrunch, 2016
- AI Server Cost Analysis – Memory Is The Biggest Loser (Patel and Wong, 29 May 2023) SemiAnalysis, 2023
- AI Server System BOM & Rack Cost: Historical Data & Chart (newest print March 2026) Silicon Analysts, 2026