Used & Refurbished Data Center GPU Market Size, Share,
Dublin, Oct. 27, 2026 (GLOBE NEWSWIRE) -- "Used & Refurbished Data Center GPU Market Size, Share, and Industry Analysis - Global Forecast to 2036" has been added to ResearchAndMarkets.com's offering.
The global used and refurbished data center GPU market was valued at USD 3.15 billion in 2025 and is estimated to reach USD 4.11 billion in 2026. Revenue is projected to increase to USD 21.54 billion by 2036, representing a CAGR of 18.0% during the forecast period.
Market growth is being driven by the accelerated replacement of AI infrastructure, expanding inference workloads, limited availability of new GPUs and high-bandwidth memory, and rising demand for cost-effective computing capacity. GPU-dense AI servers can be refreshed within 10 to 18 months, compared with four to six years for conventional enterprise equipment. NVIDIA's shift to an annual architecture cadence, progressing from Hopper to Blackwell, Blackwell Ultra, and Vera Rubin, is also increasing the volume of capable hardware entering secondary channels.
The report evaluates market dynamics, GPU lifecycle and depreciation, secondary pricing, value chain development, export controls, regulatory requirements, Porter's five forces, competitive activity, and long-term growth opportunities. It also provides forecasts and segment-level analysis across five regions and 23 countries and sub-regions.
Market Dynamics
Faster fleet refresh cycles among hyperscalers, neoclouds, enterprises, research institutions, and AI companies are creating a growing supply of Ampere and Hopper hardware with substantial remaining operational value. Demand is supported by AI inference, fine-tuning, HPC, scientific computing, visualization, VDI, and rendering workloads that do not always require the latest GPU generation. Immediate availability and lower acquisition costs further strengthen the appeal of secondary data center GPUs.
Key restraints include sharp price adjustments following new architecture launches, uncertain residual values, sustained-load failure risks, inconsistent condition grading, and limited warranty or software support. U.S. export controls covering resale and re-export also affect international transactions. In addition, liquid-cooled rack-scale systems can require specialized power, cooling, and facility infrastructure.
Growth opportunities are emerging through certified refurbishment, standardized testing and grading, sovereign AI investment, GPU-backed financing, and greater secondary price transparency. GPU price indices, compute exchanges, and futures are supporting more reliable valuations, while hyperscalers are expanding circular data center programs with ITAD partners. The market is also moving beyond individual accelerator cards toward complete GPU servers, HGX systems, compute trays, SXM/OAM modules, and rack-scale infrastructure.
Segment Outlook
By GPU architecture, NVIDIA Ampere is expected to hold the largest market share in 2026 because of the substantial installed base of A100 GPUs being replaced by Hopper and Blackwell systems. NVIDIA Blackwell & Later, including Rubin, is projected to record the fastest growth and become the largest architecture segment by 2036. Hopper is expected to surpass Ampere by 2030. Other covered architectures include NVIDIA Volta, Turing & Earlier; AMD Instinct; Ada Lovelace L40S and L4; Intel Gaudi; and Data Center GPU Max.
Refurbished GPUs are forecast to lead the condition segment in 2026 and grow faster than used equipment. Production AI buyers increasingly favor tested hardware supported by condition documentation and warranty coverage.
Complete GPU servers and systems are expected to account for the largest form factor share in 2026, reflecting the retirement of complete HGX servers and the higher transaction value of integrated systems. SXM/OAM modules and HGX baseboards are projected to grow the fastest as Hopper and Blackwell modules and compute trays enter resale channels.
ITAD and lifecycle service providers are expected to remain the leading sales channel as major fleet operators rely on certified partners for decommissioning, data sanitization, compliance, and resale. Online marketplaces and GPU exchanges are forecast to achieve the fastest growth due to improving price discovery and broader access to available inventory.
AI inference is expected to be the largest and fastest-growing application segment in 2026, supported by rising token volumes and the suitability of previous-generation GPUs for serving trained models. Neoclouds and GPU cloud providers are projected to lead among end users, while enterprises are expected to grow the fastest as private AI inference deployments increase demand for certified secondary hardware.
Regional Market Analysis
North America is expected to hold the largest global market share in 2026. The U.S. remains the leading country market because of its extensive hyperscaler and neocloud fleets, established ITAD ecosystem, active reseller network, and export regulations that favor qualified domestic buyers.
Asia-Pacific is projected to rank as the second-largest region, led by China. Regional growth is moderated by U.S. export controls and increasing adoption of domestically produced accelerators. India is expected to be the fastest-growing country market in Asia-Pacific, supported by the IndiaAI Mission and growing demand for affordable AI compute infrastructure.
The Middle East & Africa is forecast to post the fastest regional growth through 2036 as AI data center investment accelerates. The UAE is expected to remain the region's largest market, while Saudi Arabia records the highest growth. Germany leads Europe, with the Nordics expected to expand the fastest. Brazil is the largest Latin American market, while Mexico is projected to grow at a faster rate. Canada also presents opportunities through its Sovereign AI Compute Strategy.
Competitive Landscape