Processing-in-Memory Market Size, Share & Industry Analysis, By Product Type (PIM-enabled Memory ICs, PIM Memory Modules, PIM Accelerator Cards & Boards, and PIM Semiconductor IP), By Processing Technology (Digital PIM and Analog & Mixed Signal PIM), By Application (AI & ML, High-Performance Computing, Data Analytics & Database Processing, Graph & Recommendation Processing, Signal & Image Processing, and Others), By Industry (Consumer Electronics, IT & Telecom, Automotive, Healthcare, and Others), and Regional Forecast, 2026 – 2034

The global processing-in-memory market size was valued at USD 623.3 million in 2025. The market is projected to grow from USD 830.8 million in 2026 to USD 8,805.8 million by 2034, exhibiting a CAGR of 34.3% during the forecast period.

The processing-in-memory sector refers to semiconductor memory technologies that involve the placement of computational tasks within or close to the memory, hence minimizing the need to transfer data between the processors and memory. The use of such technologies greatly enhances processing speed, energy efficiency, and overall performance of data processing operations. The need for processing in memory systems is growing due to the increasing use of AI, high-performance computing, and applications with big data.

Furthermore, many key market players, such as Samsung Electronics Co., Ltd., SK hynix Inc., Hangzhou Zhicun (Witmem) Technology Co., Ltd., UPMEM SAS, and d-Matrix Corporation, operating in the market, are focusing on product innovation and strategic partnerships to develop advanced PIM architectures and accelerate their adoption across AI, HPC, and data center applications.

IMPACT OF GENERATIVE AI

Generative AI Expansion Strengthening Demand for High-Bandwidth PIM Solutions

The explosive development of generative AI is leading to a search for processing in memory solutions due to the enormous memory bandwidth and massive data movement between memory and processors required by large language models. PIM architectures work by performing part of the calculation in the memory area, thus reducing latency and increasing energy efficiency. With the increase in cases of generative AI inference being implemented in data centers, there are new chances for PIM to act in tandem with a GPU since it can process many memory-heavy applications including the attention mechanism. Memory challenges arise as Ai models enable longer context and higher user concurrency. Thus, generative AI can be seen as a fast development and increasing application of PIM. For instance,

• In October 2025, SK hynix showcased its enhanced AiM-based AiMX Processing in Memory solution at the AI Infra Summit 2025, demonstrating a system combining NVIDIA H100 GPUs with AiMX cards for large language model inference.

Thus, this factor fuels the processing-in-memory market growth in the coming years.

PROCESSING-IN-MEMORY MARKET TRENDS

Rising Investment in Analog Compute-in-Memory for Edge and Energy-Constrained AI Workloads Fuels Market Growth

The increase in investment in analog computing in memory is boosting the market, since AI applications are requiring more computing power in less energy. Analog compute in memory is running selected AI processes inside memory chips, thus minimizing the need to transfer data between memory and processors and making the process more energy-efficient. Such architecture is ideal for edge AI functions such as robotics, autonomous systems, smart devices, industrial equipment, and defense systems, where energy, space, and cooling capacity are limited. Since generative AI applications and inference tasks move to the edge, companies are increasingly motivated to invest in analog compute in memory development. For instance,

• In February 2025, EnCharge AI raised more than USD 100 million in Series B funding, bringing its total funding to over USD 144 million, to commercialize its analog in-memory computing AI accelerator products.

MARKET DYNAMICS

MARKET DRIVERS

To get more information on the regional analysis of this market, Download Free sample

AI Inference Growth and Memory-Bandwidth Bottlenecks Drives Market Growth

The demand for PIM solutions is increasing due to rapid growth in the use of Artificial Intelligence (AI) inference. Large Language Models require an uninterrupted use of model weights and key-value cache data, making it difficult for traditional processor-memory systems, which have the potential of hitting the problems concerning memory bandwidth. The performance in standard processor-memory systems usually suffers from constant back and forth movement of data between DRAM and GPUs or accelerators. PIM performs specific computations already in the memory, which helps to cut down the data movement, and improves the effective bandwidth, latency and energy efficiency as a result. The importance of these benefits grows with the increasing use of generative AI solutions, for which larger models, longer context windows and higher number of users are required. Thus, rising AI inference workloads contribute to the development of PIM architectures by semiconductor manufacturers. For instance,

• August 2026, Samsung Electronics demonstrated its LPDDR5X-PIM technology at Hot Chips 2026, specifically targeting AI inference workloads by integrating computing functions directly within low-power DRAM.

MARKET RESTRAINTS

Long Qualification Cycles, Software Ecosystem Immaturity and Integration Complexity May Hinder Market Growth

PIM solutions are slow to be adopted because semiconductor customers typically demand long qualification and validation processes prior to integrating new memory architectures into their commercial systems. The software ecosystem of PIM is also less mature than traditional CPU and GPU ecosystems.

PIM integration can be very complicated due to its dependencies on memory hardware, processors, system architecture, and software frameworks. Hence, qualification delays and ecosystem issues may hinder the process of commercialization of PIM technologies.

MARKET OPPORTUNITIES

Rising Demand for Energy-Efficient Data Centers to Create New Opportunities for Market Growth

Rising power consumption in AI and high-performance computing data centers is creating a strong opportunity for processing-in-memory solutions. Existing systems consume power for massive data transfers between processors and memory systems/disk, while PIM minimizes such operations by carrying out specific calculations that are performed close to or in memory itself. This method works in the direction of boosting perf/W, lowering latency, and cutting power consumption. As data center owners are paying more attention to minimizing the cost of electricity and increasing efficiency of the infrastructure, PIM could become a good supplement to systems based on GPU. With growing applications of generative AI and enormous language models, the prospects for energy-efficient PIM systems are increasing.

Segmentation Analysis

By Product Type

Growing AI and HPC Workloads Strengthened PIM-Enabled Memory IC Segment Adoption

Based on product type, the market is categorized into PIM-enabled memory ICs, PIM memory modules, PIM accelerator cards & boards, and PIM semiconductor IP.

PIM-enabled memory ICs segment accounted for the largest processing-in-memory market share in 2025. This is owing to their ability to combine processing capabilities with memory itself which helps to minimize data movement and enhance utilization of bandwidth, latency, and energy efficiency. The increasing application of PIM-enabled memory ICs in AI inference, high-performance computing, and data center sectors only fueled the demand, since those workloads need memory architecture which is faster and requires less power.

PIM semiconductor IP segment is anticipated to grow at the highest CAGR of 39.9% in the coming years. This is owing to increasing demand from chipmakers for customizable, licensable PIM architectures that can be integrated into AI accelerators, memory devices, and edge computing platforms while reducing development time and cost.

By Processing Technology

Higher Programmability and System Compatibility Boosted Digital PIM Segment Dominance

Based on processing technology, the market is bifurcated into digital PIM and analog & mixed-signal PIM.

Digital PIM segment accounted for the largest market share in 2025. This is owing to its ability to integrate seamlessly with existing systems based on digital memory architecture, semiconductor design processes, and conventional computing system make it the number one market leader. Additionally, its capacity to offer predictable accuracy and easier re-programming capability despite its deployment across HPC, data center, and AI inference jobs strengthened its commercial uptake compared to other analog approaches to PIM which are less advanced.

Analog & mixed signal PIM segment is anticipated to grow at the highest CAGR of 38.6% over the forecast period. This is owing to its strong potential to perform AI computations with significantly lower data movement and power consumption, making it increasingly suitable for edge AI, embedded systems, and energy-constrained inference workloads.

By Application

High Memory-Bandwidth Requirements Supported AI & ML Segment Leadership

Based on application, the market is divided into AI & ML, high-performance computing, data analytics & database processing, graph & recommendation processing, signal & image processing, and others (encapsulation, potting, sealing, etc.).

Artificial Intelligence and Machine Learning (AI & ML) segment accounted for the largest market share in 2025. This is owing to the rising demand for large language models, recommendation systems, computer vision, and other data-heavy applications. PIM designs can minimize the transfer of data between processors and memory, which helps enhance performance and efficiency, making them perfect candidates for AI-related training and inference tasks.

Others segment is anticipated to grow at the highest CAGR of 39.9% over the forecast period. This is owing to increasing demand for advanced protective materials that provide electrical insulation, moisture resistance, thermal stability, and mechanical protection for increasingly compact and high-performance electronic components.

To know how our report can help streamline your business, Speak to Analyst

By Industry

Generative AI and 5G Expansion Boosted PIM Demand in IT & Telecom Segment

Based on industry, the market is classified into consumer electronics, IT & telecom, automotive, healthcare, and others (government, manufacturing, etc.).

IT & Telecom segment witnessed a dominating market share in 2025. This is owing to the numerous AI servers, data centers, cloud infrastructure, and networking systems which necessitate low latency processing and high memory bandwidth. Another contributor to the growth in demand for PIM is the heightened use of generative AI, 5G technology, hyperscale data facilities, and massive data analytics in the sector.

Automotive segment is anticipated to grow at the highest CAGR of 37.8% during the forecast period. This is owing to the rising integration of ADAS, autonomous driving, in-vehicle AI, and real-time sensor data processing systems that require low-latency, high-bandwidth, and energy-efficient computing architectures such as PIM.

Processing-in-Memory Market Regional Outlook

By geography, the market is categorized into North America, South America, Europe, the Middle East & Africa, and Asia Pacific.

Asia Pacific

To get more information on the regional analysis of this market, Download Free sample

Asia Pacific held the largest market share in 2024, valuing at USD 210.2 million, and also maintained the leading share in 2025, with USD 280.8 million. The market in Asia Pacific is expected to grow at the highest CAGR of 35.6% during the forecast period, owing to its robust semiconductor manufacturing ecosystem plus presence of well-known companies in the domain of memory technology, including Samsung Electronics and SK Hynix, which are working on PIM technologies. Increasing investments in AI infrastructure along with high performance computing and advanced memory as well as data centers in countries including South Korea, China, Japan, and Taiwan fostered the regional market demand for high bandwidth energy-efficient memory architectures. In the region, India and China are both estimated to reach USD 39.5 million and USD 105.2 million, respectively, in 2026. For instance,

• In January 2025, South Korea-based SK hynix showcased its next-generation PIM technologies at CES 2025 as part of its “Full Stack AI Memory Provider” strategy, alongside HBM, CXL, server DRAM, and on-device AI memory solutions. The development highlights continued investment by APAC-based memory manufacturers in PIM and other advanced AI-memory technologies.

These factors play a significant role in fueling the market growth.

China Processing-in-Memory Market

China’s market is projected to be one of the largest globally, with 2026 revenues estimated at around USD 105.2 million, representing roughly 16.9% of global sales. This is owing to rapid expansion of domestic AI infrastructure, data centers, semiconductor manufacturing, and government-backed advanced computing initiatives that are increasing demand for high-bandwidth, energy-efficient memory architectures.

Japan Processing-in-Memory Market

The Japanese market is estimated at around USD 50.7 million in 2026, accounting for roughly 8.1% of global revenues.

India Processing-in-Memory Market

The Indian market is estimated at around USD 39.5 million in 2026, accounting for roughly 6.3% of global revenues.

North America

North America is estimated to reach USD 228.1 million in 2026 and secure the position of the second-largest region in the market. This is owing to the proliferation of generative AI applications, increasing number of hyperscale data centers and high-performance computing infrastructure which requires memory architectures to save bandwidth and reduce energy consumption. Other factors driving digital in-memory computing and memory-centric AI accelerators are advancements in semiconductor technologies, availability of venture capital as well as major companies in the US working in cloud computing and AI sector. For instance,

• In November 2025, U.S.-based d-Matrix raised USD 275 million in Series C funding to expand its AI inference platform based on compute-memory integration and accelerate large-scale deployments across hyperscale, enterprise, and sovereign data centers.

U.S. Processing-in-Memory Market

Based on North America’s significant contribution and the U.S. dominance within the region, the U.S. market can be analytically approximated at around USD 189.3 million in 2026, accounting for roughly 30.4% of global processing-in-memory sales.

Europe

Europe is projected to record a growth rate of 33.4% in the coming years, which is the third highest among all regions, and reach a valuation of USD 155.6 million by 2026. The market growth in Europe is increasing, owing to the growing investments in government AI infrastructure, energy-efficient semiconductor technologies, and high-performance computing in different nations including Germany, France, the Netherlands, Belgium, and the U.K. The rise in the market for edge AI and generative AI inference is also aiding the expansion of PIM and compute-in-memory architecture, aimed at minimizing power consumption, data movement, and bandwidth limitations. For instance,

• In March 2025, Netherlands-based Axelera AI unveiled Titania, a high-performance AI inference chiplet based on its Digital In-Memory Computing (D-IMC) architecture, and secured up to USD 67 million in funding from the EuroHPC Joint Undertaking and participating member states under the DARE project.

U.K. Processing-in-Memory Market

The U.K. market is estimated at around USD 24.1 million in 2026, representing roughly 3.9% of global revenues.

Germany Processing-in-Memory Market

Germany’s market is projected to reach approximately USD 30.3 million in 2026, equivalent to around 4.9% of global sales.

South America

South America is expected to witness moderate growth in this market during the forecast period. The market is set to reach a valuation of USD 27.1 million in 2026. This is owing to increasing investments in data centers, cloud infrastructure, AI adoption, and telecom modernization, particularly in Brazil, which are creating demand for faster and more energy-efficient memory processing solutions.

Brazil Processing-in-Memory Market

The Brazilian market is set to reach a value of USD 15.2 million in 2026.

Middle East & Africa

The Middle East & Africa is estimated to reach USD 44.3 million in 2026 and is expected to grow at a prominent growth rate in the coming years. This is owing to increasing investments in AI infrastructure, hyperscale data centers, cloud computing, and sovereign digital transformation programs, particularly across the GCC countries. Also, increasing development in demand for enhanced analytics, smart city technology, modernization of telecommunications, and energy saving computing also boosts opportunities for high-speed, low-stop memory systems such as the PIM where the input of information can take place instantaneously.

GCC Processing-in-Memory Market

The GCC market is set to reach a value of USD 14.2 million in 2026.

COMPETITIVE LANDSCAPE

Key Industry Players

Focus on AI Acceleration, Memory Bandwidth Optimization, and Energy-Efficient In-Memory Computing by Key Players to Propel Market Growth

The global processing-in-memory market holds a semi-consolidated market structure, with prominent players such as Samsung Electronics Co., Ltd., SK hynix Inc., Hangzhou Zhicun (Witmem) Technology Co., Ltd., UPMEM SAS, and d-Matrix Corporation. These companies are focusing on advanced PIM architectures, AI memory accelerators, high-bandwidth memory integration, and compute-in-memory solutions to reduce data movement, improve processing speed, and lower power consumption across AI inference, high-performance computing, data centers, edge computing, and enterprise workloads.

Other notable players in the global market include Axelera AI B.V., Mythic, Inc., GSI Technology, Inc., TetraMem Inc., and EnCharge AI, Inc. Growing competition around digital and analog in-memory computing, memory-centric AI acceleration, semiconductor IP, and energy-efficient inference technologies is expected to accelerate product innovation, strengthen the software ecosystem, and expand PIM adoption across generative AI, autonomous systems, robotics, telecommunications, automotive electronics, and next-generation data center infrastructure.

LIST OF KEY PROCESSING-IN-MEMORY COMPANIES PROFILED IN REPORT

• Samsung Electronics Co., Ltd. (South Korea)
• SK hynix Inc. (South Korea)
• Hangzhou Zhicun (Witmem) Technology Co., Ltd. (China)
• UPMEM SAS (France)
• d-Matrix Corporation (U.S.)
• Axelera AI B.V. (Netherlands)
• Mythic, Inc. (U.S.)
• GSI Technology, Inc. (U.S.)
• TetraMem Inc. (U.S.)
• EnCharge AI, Inc. (S.)

KEY INDUSTRY DEVELOPMENTS

• February 2026: Mythic and Honda R&D announced a joint development agreement to create an automotive-grade AI SoC using Mythic's analog compute-in-memory architecture for Honda's next-generation software-defined vehicles. Honda will license Mythic's Analog Processing Unit technology, with the companies targeting highly energy-efficient AI processing for ADAS, autonomous driving, and other onboard AI workloads.
• January 2026: GSI Technology announced a government-funded proof-of-concept engagement with G2 Tech to develop the Sentinel autonomous perimeter-security platform using its Gemini-II compute-in-memory APU. The collaboration applies Gemini-II to real-time drone and camera processing, demonstrating the commercial potential of compute-in-memory technology for low-power edge AI applications.
• May 2025: EnCharge AI launched the EN100 AI accelerator, built on its precise and scalable analog in-memory computing technology for laptops, workstations, and edge devices. The accelerator provides more than 200 TOPS of AI compute performance while targeting substantially improved performance per watt for on-device inference workloads.
• November 2025: d-Matrix and Alchip announced a collaboration to develop a 3D DRAM-based data center inference accelerator built around d-Matrix’s Digital In-Memory Compute architecture. Their 3DIMC technology is planned for commercial deployment in d-Matrix’s Raptor accelerator and is intended to reduce memory-related performance and cost bottlenecks in large-scale AI inference.
• October 2025: Axelera AI announced its Europa AIPU, featuring eight second-generation AI cores based on the company’s Digital In-Memory Compute technology and delivering up to 629 TOPS at INT8 precision. The accelerator is designed for generative AI and computer vision inference from edge systems to enterprise servers while addressing memory bandwidth, power efficiency, and processing-cost requirements.
• December 2024: UPMEM selected Semidynamics’ RISC-V AI processor IP for its next-generation LPDDR5X Processing-in-Memory device aimed at generative AI and large language model applications.
• November 2024: TetraMem and SK hynix announced a joint research partnership focused on advancing in-memory computing technologies for artificial intelligence The collaboration combines TetraMem's analog in-memory computing platform with SK hynix's memory expertise to improve the scalability, processing efficiency, and commercial accessibility of memory-centric AI architectures.

REPORT COVERAGE

The global processing-in-memory market analysis includes a comprehensive study of the market size & forecast by all the market segments included in the report. It includes details on the market dynamics and market trends expected to drive the market over the forecast period. It provides information on key aspects, including an overview of technological advancements, pipeline candidates, the regulatory environment, and product launches. Additionally, it details partnerships, mergers & acquisitions, as well as key industry developments. The global market research report also provides a detailed competitive landscape with information on the market share and profiles of key operating players.

Request for Customization   to gain extensive market insights.

Report Scope & Segmentation

Research Methodology

1. Bottom Up Approaches

Approach 1: Revenues and Market Share of Major PIM Players, 2021–2025

• Collection of PIM-specific and attributable revenues from annual reports, investor presentations, company disclosures, product portfolios, and commercialization announcements of companies such as Samsung Electronics, SK hynix, UPMEM, Witmem, d-Matrix, Axelera AI, GSI Technology, TetraMem, EnCharge AI, and others.
• Company-level PIM revenue was estimated based on commercial product availability, shipment volumes, customer deployments, product pricing, production status, and attributable PIM content.
• The estimated revenues of major suppliers were aggregated and adjusted for smaller and emerging vendors to derive the global market size.

Approach 2: Unit Shipment and ASP Analysis, 2021–2025

• Estimation of annual shipments of PIM-enabled memory ICs, PIM memory modules, accelerator cards & boards, and commercially monetized PIM semiconductor IP.
• Product volumes were multiplied by estimated weighted average selling prices based on memory type, processing capability, packaging, product configuration, and commercialization stage.
• The summation of product-level revenues was used to estimate and validate the overall processing-in-memory market.

Approach 3:  Application and Workload Demand Model, 2021–2025

• Assessment of PIM adoption across AI & ML, HPC, database and analytics processing, graph and recommendation processing, signal and image processing, and other memory-intensive workloads.
• Addressable computing infrastructure was evaluated based on memory bandwidth requirements, data-movement intensity, processing efficiency, hardware availability, and commercial PIM penetration.
• Application-level demand was aggregated across regions and reconciled with supplier revenues and shipment estimates.

2. Top Down Approaches

Approach 1: Analysis of Parent and Addressable Semiconductor Markets, 2021–2025

• Analysis of relevant parent markets including DRAM, HBM, SRAM, emerging memory, AI accelerator hardware, and high-performance computing infrastructure.
• The technically addressable portion of these markets was identified and suitable PIM penetration assumptions were applied based on commercial readiness, memory bandwidth requirements, AI adoption, and regional semiconductor activity.
• The resulting estimate was cross-validated with company revenues, product shipments, and commercial PIM deployments to avoid overestimation.

3. List of Key Sources

• World Semiconductor Trade Statistics (WSTS)
• Semiconductor Industry Association (SIA)
• JEDEC
• SEMI
• IEEE
• Global Semiconductor Alliance (GSA)
• CXL Consortium
• Annual reports
• investor presentations
• SEC filings
• Earnings releases
• Product datasheets
• Technical papers
• Developer documentation
• Commercialization announcements of PIM companies
• U.S. Department of Commerce
• BIS
• European Commission
• METI Japan
• MOTIE South Korea
• Taiwan MOEA
• MeitY India
• OECD
• World Bank
• IEA
• ITU
• Other relevant institutional sources

4. Primary Interviews

Supply Side Interviews: 60%

• Key Respondents: PIM chip developers, memory manufacturers, accelerator vendors, semiconductor IP companies, module manufacturers, foundries, packaging companies, distributors, and system suppliers.
• Interviews were used to validate shipment volumes, ASP ranges, commercialization timelines, product mix, regional demand, and competitive positioning.

Demand Side Interviews: 40%

• Key Respondents: Hyperscalers, cloud providers, data-center operators, HPC users, AI infrastructure providers, OEMs, system integrators, and enterprise technology buyers.
• Interviews were used to assess PIM adoption, workload suitability, purchasing preferences, deployment patterns, performance requirements, and future adoption across countries and regions.