The global Used & Refurbished Data Center GPU Market was valued at USD 3.15 billion in 2025. It is estimated to reach USD 4.11 billion in 2026 and is projected to grow to USD 21.54 billion by 2036, at a CAGR of 18.0% during the forecast period. This report offers a detailed look at the market, covering market dynamics, GPU lifecycle and depreciation, secondary pricing, the ecosystem and value chain, export controls and other regulation, Porter's five forces, competitive activity, and future growth opportunities.
Used and refurbished data center GPUs are graphics processing units and GPU-based systems that have completed a first period of service in a data center and are sold to a second owner. They include PCIe accelerator cards, SXM and OAM modules, HGX baseboards, and complete GPU servers and systems, sourced mainly from hyperscalers, GPU cloud providers, enterprises, research centers, and AI startups that retire or rotate their fleets. The market measures the final sale value paid by end buyers and excludes new GPUs, consumer GPUs, cloud GPU rental, ITAD service fees, scrap value, dealer-to-dealer trades, and untraceable grey-market flows.
AI hardware is being replaced far faster than conventional IT. GPU-dense AI servers are refreshed in as little as 10 to 18 months, against four to six years for traditional enterprise equipment, and NVIDIA has moved to a yearly architecture cadence from Hopper to Blackwell, Blackwell Ultra, and Vera Rubin. At the same time, inference workloads are growing rapidly and run well on GPUs that are no longer used for frontier training, while shortages of new GPUs and high-bandwidth memory have made immediately available secondary hardware more valuable.
The report examines market drivers, restraints, opportunities, challenges, and trends, the GPU lifecycle and depreciation policies of hyperscalers and neoclouds, secondary prices by generation, condition, and channel, the value chain, U.S. export controls and enforcement, e-waste and data sanitization rules, and Porter's five forces. It explains why NVIDIA Blackwell & Later, SXM/OAM modules and HGX baseboards, online marketplaces and GPU exchanges, and enterprises are growing faster than the overall market. It also provides market forecasts, segment-level insights, and regional analysis across 5 regions and 23 countries and sub-regions to support business, investment, and product decisions.
Secondary GPU prices fall in steps at each new architecture launch, which makes residual values uncertain and creates inventory risk for resellers. Failure rates under sustained AI load, limited warranty and software support, and inconsistent condition grading also restrain growth. In addition, U.S. export controls that apply to resale and re-export, and the power, cooling, and facility requirements of liquid-cooled rack-scale systems, remain challenges.
Certified refurbishment, testing, and grading programs, demand from sovereign AI programs and emerging markets, and GPU-backed financing that depends on reliable residual values offer significant opportunities. Meanwhile, GPU price indices, compute exchanges, and futures are bringing transparency to secondary prices, hyperscalers are expanding circular data center programs with ITAD partners, and secondary trade is shifting from individual cards toward complete servers and rack-scale systems.
Based on GPU architecture, the market is segmented into NVIDIA Ampere; NVIDIA Hopper; NVIDIA Volta, Turing & Earlier; NVIDIA Blackwell & Later, including Rubin; AMD Instinct; and other architectures, including Ada Lovelace L40S and L4 and Intel Gaudi and Data Center GPU Max. In 2026, NVIDIA Ampere is expected to account for the largest share of the market, reflecting the large installed base of A100 GPUs being replaced by Hopper and Blackwell. NVIDIA Blackwell & Later is expected to grow the fastest from a small base and to become the largest segment by 2036, while Hopper is expected to overtake Ampere by 2030.
Based on condition, the market is segmented into refurbished and used. In 2026, refurbished GPUs are expected to hold the larger share and to grow faster, as buyers running production AI workloads favor tested units with documentation and warranty.
Based on form factor, the market is segmented into complete GPU servers and systems, SXM/OAM modules and HGX baseboards, and PCIe cards. In 2026, complete GPU servers and systems are expected to hold the largest share, reflecting the decommissioning of complete HGX servers and the higher value of full systems. SXM/OAM modules and HGX baseboards are expected to grow the fastest, driven by the resale of Hopper and Blackwell modules and compute trays.
Based on sales channel, the market is segmented into ITAD and lifecycle service providers, independent resellers and brokers, direct fleet sales, and online marketplaces and GPU exchanges. In 2026, ITAD and lifecycle service providers are expected to hold the largest share, as hyperscalers sell retired fleets through certified partners. Online marketplaces and GPU exchanges are expected to grow the fastest, supported by new exchanges and better price transparency.
Based on application, the market is segmented into AI inference; AI training and fine-tuning; HPC and scientific computing; visualization, VDI, and rendering; and other applications. In 2026, AI inference is expected to hold the largest share and to grow the fastest, driven by rapidly rising token volumes and the suitability of older GPUs for serving trained models.
Based on end user, the market is segmented into neoclouds and GPU cloud providers, enterprises, AI startups and model developers, research and academic institutions, government and defense, and other end users. In 2026, neoclouds and GPU cloud providers are expected to hold the largest share, given their need for low-cost capacity. Enterprises are expected to grow the fastest, driven by private AI inference and the lower cost of certified secondary hardware.
In 2026, North America is expected to account for the largest share of the global market, with the U.S. as the largest country market in the world, supported by the largest hyperscaler and neocloud fleets, the largest cluster of ITAD providers and resellers, and export rules that favor domestic buyers. Asia-Pacific is expected to be the second-largest region, with China as its largest market, although its growth is moderated by U.S. export controls and the rising use of domestic accelerators. India is expected to be the fastest-growing country market in the region, supported by the IndiaAI Mission and rising demand for affordable compute.
The Middle East & Africa is expected to grow the fastest during the forecast period, driven by rapidly growing AI data center capacity, with the UAE as the largest market in the region and Saudi Arabia as the fastest-growing. In Europe, Germany is the largest market and the Nordics are expected to grow the fastest; in North America, the U.S. is the largest market and Canada is supported by its Sovereign AI Compute Strategy; and in Latin America, Brazil is the largest market and Mexico is expected to grow faster.
The market is highly fragmented, with competition among large ITAD and lifecycle service providers, independent resellers and brokers, and new GPU marketplaces. Based on 2025 market share, Iron Mountain ranked first, followed by Sims Lifecycle Services, SK tes, Procurri, Alta Technologies, and Big Data Supply, while a large number of small resellers, brokers, and fleet owners account for most sales. Acquisitions and expansions are the main strategies, followed by product launches, partnerships, and agreements, including Iron Mountain's acquisitions of Wisetek, ACT Group, and Group ATF, SK tes's new facilities in Virginia and Ireland, Compute Exchange's launch of a secondary GPU marketplace in July 2026, and SK tes's memorandum of understanding with XTREME-D to offer cloud services on refurbished AI servers in Japan.
The competitive benchmarking compares companies on ITAD and decommissioning, certified data sanitization, complete GPU servers, SXM/OAM modules and HGX boards, PCIe GPU cards, online marketplaces, testing and warranty, and global footprint. Companies compete on access to hyperscale and neocloud supply, testing and certification, warranty terms, provenance and export compliance, pricing, and speed of delivery.
Key companies profiled in the report include Iron Mountain Incorporated (U.S.), Sims Limited (Australia), TES-AMM (Singapore) Pte. Ltd. (SK tes) (Singapore), Procurri Corporation Limited (Singapore), Alta Technologies, Inc. (U.S.), Big Data Supply, Inc. (U.S.), The Compute Exchange Inc. (U.S.), server-parts.eu Kft. (Hungary), exIT Technologies (U.S.), NewServerLife LLC (U.S.), BrightStar Systems (U.S.), LA Micro Group (UK) Limited (U.K.), and PC Server & Parts (U.S.).
Used and refurbished data center GPUs are graphics processing units and GPU-based systems that have completed a first period of service in a data center and are sold to a second owner. They include PCIe accelerator cards, SXM and OAM modules, HGX baseboards, and complete GPU servers and systems, sourced mainly from hyperscalers, GPU cloud providers, enterprises, research centers, and AI startups that retire or rotate their fleets. The market measures the final sale value paid by end buyers and excludes new GPUs, consumer GPUs, cloud GPU rental, ITAD service fees, scrap value, dealer-to-dealer trades, and untraceable grey-market flows.
AI hardware is being replaced far faster than conventional IT. GPU-dense AI servers are refreshed in as little as 10 to 18 months, against four to six years for traditional enterprise equipment, and NVIDIA has moved to a yearly architecture cadence from Hopper to Blackwell, Blackwell Ultra, and Vera Rubin. At the same time, inference workloads are growing rapidly and run well on GPUs that are no longer used for frontier training, while shortages of new GPUs and high-bandwidth memory have made immediately available secondary hardware more valuable.
The report examines market drivers, restraints, opportunities, challenges, and trends, the GPU lifecycle and depreciation policies of hyperscalers and neoclouds, secondary prices by generation, condition, and channel, the value chain, U.S. export controls and enforcement, e-waste and data sanitization rules, and Porter's five forces. It explains why NVIDIA Blackwell & Later, SXM/OAM modules and HGX baseboards, online marketplaces and GPU exchanges, and enterprises are growing faster than the overall market. It also provides market forecasts, segment-level insights, and regional analysis across 5 regions and 23 countries and sub-regions to support business, investment, and product decisions.
Market Dynamics
Faster refresh cycles of AI GPU fleets are the main force behind market growth, as Hopper and Ampere systems leave first-owner fleets years before the end of their useful life. The cost advantage and immediate availability of secondary GPUs during shortages of new GPUs and memory, together with the rapid growth of inference workloads and neocloud capacity, reinforce demand.Secondary GPU prices fall in steps at each new architecture launch, which makes residual values uncertain and creates inventory risk for resellers. Failure rates under sustained AI load, limited warranty and software support, and inconsistent condition grading also restrain growth. In addition, U.S. export controls that apply to resale and re-export, and the power, cooling, and facility requirements of liquid-cooled rack-scale systems, remain challenges.
Certified refurbishment, testing, and grading programs, demand from sovereign AI programs and emerging markets, and GPU-backed financing that depends on reliable residual values offer significant opportunities. Meanwhile, GPU price indices, compute exchanges, and futures are bringing transparency to secondary prices, hyperscalers are expanding circular data center programs with ITAD partners, and secondary trade is shifting from individual cards toward complete servers and rack-scale systems.
Segment Analysis
The report analyzes the market by GPU architecture, condition, form factor, sales channel, application, end user, and geography, helping stakeholders identify the most promising growth opportunities.Based on GPU architecture, the market is segmented into NVIDIA Ampere; NVIDIA Hopper; NVIDIA Volta, Turing & Earlier; NVIDIA Blackwell & Later, including Rubin; AMD Instinct; and other architectures, including Ada Lovelace L40S and L4 and Intel Gaudi and Data Center GPU Max. In 2026, NVIDIA Ampere is expected to account for the largest share of the market, reflecting the large installed base of A100 GPUs being replaced by Hopper and Blackwell. NVIDIA Blackwell & Later is expected to grow the fastest from a small base and to become the largest segment by 2036, while Hopper is expected to overtake Ampere by 2030.
Based on condition, the market is segmented into refurbished and used. In 2026, refurbished GPUs are expected to hold the larger share and to grow faster, as buyers running production AI workloads favor tested units with documentation and warranty.
Based on form factor, the market is segmented into complete GPU servers and systems, SXM/OAM modules and HGX baseboards, and PCIe cards. In 2026, complete GPU servers and systems are expected to hold the largest share, reflecting the decommissioning of complete HGX servers and the higher value of full systems. SXM/OAM modules and HGX baseboards are expected to grow the fastest, driven by the resale of Hopper and Blackwell modules and compute trays.
Based on sales channel, the market is segmented into ITAD and lifecycle service providers, independent resellers and brokers, direct fleet sales, and online marketplaces and GPU exchanges. In 2026, ITAD and lifecycle service providers are expected to hold the largest share, as hyperscalers sell retired fleets through certified partners. Online marketplaces and GPU exchanges are expected to grow the fastest, supported by new exchanges and better price transparency.
Based on application, the market is segmented into AI inference; AI training and fine-tuning; HPC and scientific computing; visualization, VDI, and rendering; and other applications. In 2026, AI inference is expected to hold the largest share and to grow the fastest, driven by rapidly rising token volumes and the suitability of older GPUs for serving trained models.
Based on end user, the market is segmented into neoclouds and GPU cloud providers, enterprises, AI startups and model developers, research and academic institutions, government and defense, and other end users. In 2026, neoclouds and GPU cloud providers are expected to hold the largest share, given their need for low-cost capacity. Enterprises are expected to grow the fastest, driven by private AI inference and the lower cost of certified secondary hardware.
Regional Analysis
The report covers North America, Europe, Asia-Pacific, Latin America, and the Middle East & Africa, spanning 23 countries and sub-regions. The regional analysis considers the installed base of data center GPUs, hyperscaler and neocloud capacity, the presence of ITAD providers and resellers, sovereign AI programs, and export control rules.In 2026, North America is expected to account for the largest share of the global market, with the U.S. as the largest country market in the world, supported by the largest hyperscaler and neocloud fleets, the largest cluster of ITAD providers and resellers, and export rules that favor domestic buyers. Asia-Pacific is expected to be the second-largest region, with China as its largest market, although its growth is moderated by U.S. export controls and the rising use of domestic accelerators. India is expected to be the fastest-growing country market in the region, supported by the IndiaAI Mission and rising demand for affordable compute.
The Middle East & Africa is expected to grow the fastest during the forecast period, driven by rapidly growing AI data center capacity, with the UAE as the largest market in the region and Saudi Arabia as the fastest-growing. In Europe, Germany is the largest market and the Nordics are expected to grow the fastest; in North America, the U.S. is the largest market and Canada is supported by its Sovereign AI Compute Strategy; and in Latin America, Brazil is the largest market and Mexico is expected to grow faster.
Competitive Landscape
The report provides a detailed review of the competitive landscape, covering key growth strategies, competitive benchmarking, the competitive dashboard, and market share analysis by key player, together with profiles of leading companies including their business and financial overview, product portfolios, strategy and market position, strategic developments, and SWOT analyses.The market is highly fragmented, with competition among large ITAD and lifecycle service providers, independent resellers and brokers, and new GPU marketplaces. Based on 2025 market share, Iron Mountain ranked first, followed by Sims Lifecycle Services, SK tes, Procurri, Alta Technologies, and Big Data Supply, while a large number of small resellers, brokers, and fleet owners account for most sales. Acquisitions and expansions are the main strategies, followed by product launches, partnerships, and agreements, including Iron Mountain's acquisitions of Wisetek, ACT Group, and Group ATF, SK tes's new facilities in Virginia and Ireland, Compute Exchange's launch of a secondary GPU marketplace in July 2026, and SK tes's memorandum of understanding with XTREME-D to offer cloud services on refurbished AI servers in Japan.
The competitive benchmarking compares companies on ITAD and decommissioning, certified data sanitization, complete GPU servers, SXM/OAM modules and HGX boards, PCIe GPU cards, online marketplaces, testing and warranty, and global footprint. Companies compete on access to hyperscale and neocloud supply, testing and certification, warranty terms, provenance and export compliance, pricing, and speed of delivery.
Key companies profiled in the report include Iron Mountain Incorporated (U.S.), Sims Limited (Australia), TES-AMM (Singapore) Pte. Ltd. (SK tes) (Singapore), Procurri Corporation Limited (Singapore), Alta Technologies, Inc. (U.S.), Big Data Supply, Inc. (U.S.), The Compute Exchange Inc. (U.S.), server-parts.eu Kft. (Hungary), exIT Technologies (U.S.), NewServerLife LLC (U.S.), BrightStar Systems (U.S.), LA Micro Group (UK) Limited (U.K.), and PC Server & Parts (U.S.).
How This Report Helps
- Provides reliable market size estimates and long-term forecasts for the global used and refurbished data center GPU market.
- Explains how faster AI fleet refresh cycles, new-GPU and memory shortages, inference growth, and neocloud expansion affect market growth.
- Identifies high-growth opportunities across GPU architecture, condition, form factor, sales channel, application, end user, and region.
- Analyzes GPU lifecycle and depreciation, secondary pricing, the value chain, export controls, and data sanitization rules shaping the market.
- Benchmarks leading companies based on their capabilities and market share.
- Supports fleet disposal and procurement planning, residual value assessment, partnership and acquisition evaluation, and market entry decisions.
- Delivers practical market insights for ITAD and lifecycle service providers, refurbishers, resellers, marketplaces, hyperscalers, neoclouds, enterprises, lenders, and investors.
Key Questions Answered
- How big is the global used and refurbished data center GPU market today, and how will it grow through 2036?
- How are faster refresh cycles, GPU and memory shortages, and inference workloads driving demand?
- What are the main drivers, restraints, opportunities, and challenges shaping the industry?
- Which GPU architecture, condition, form factor, sales channel, application, end-user, and regional segments will grow the fastest?
- How quickly do secondary GPU prices fall after each new architecture launch, and how do prices differ by condition and channel?
- How do U.S. export controls and data sanitization requirements affect the resale of data center GPUs?
- How are GPU price indices, compute exchanges, GPU-backed financing, and rack-scale resale changing the market?
- Which regions and countries offer the most attractive business opportunities?
- Who are the leading companies in the market, and what strategies are they following?
Table of Contents
1. Introduction
2. Research Methodology
4. Market Overview
5. Used & Refurbished Data Center GPU Market Assessment, by GPU Architecture
6. Used & Refurbished Data Center GPU Market Assessment, by Condition
7. Used & Refurbished Data Center GPU Market Assessment, by Form Factor
8. Used & Refurbished Data Center GPU Market Assessment, by Sales Channel
9. Used & Refurbished Data Center GPU Market Assessment, by Application
10. Used & Refurbished Data Center GPU Market Assessment, by End User
11. Used & Refurbished Data Center GPU Market Assessment, by Geography
12. Competitive Landscape
13. Company Profiles
14. Appendix
List of Tables
List of Figures
Companies Mentioned
- Iron Mountain Incorporated
- Sims Limited
- TES-AMM
- Procurri Corporation Limited
- Alta Technologies, Inc.
- Big Data Supply, Inc.
- The Compute Exchange Inc.
- server-parts.eu Kft.
- exIT Technologies
- NewServerLife LLC
- BrightStar Systems
- LA Micro Group
- PC Server & Parts (U.S.).
