Global Computing in Memory Technology Market Strategic Research Report

Scope of the Report

The global Computing in Memory Technology market size is predicted to grow from US$ 661 million in 2025 to US$ 7,398 million in 2032; it is expected to grow at a CAGR of 41.7% from 2026 to 2032.

As a new computing architecture, storage-computing integration is considered to be a revolutionary technology with potential and has received great attention at home and abroad. The core is to fully integrate storage and computing, effectively overcome the bottleneck of the von Neumann architecture, and combine advanced packaging and new storage devices in the post-Moore era to achieve an order of magnitude improvement in computing energy efficiency.

According to the distance between storage and computing, the technical solutions of generalized storage-computing integration are divided into three categories, namely, Processing Near Memory (PNM), Processing ln Memory (PlM) and Computing in Memory (CIM). In-memory computing is storage-computing integration in a narrow sense.

Global key players of Computing in Memory Technology include Syntiant, Zhicun(Witmem) Technology, Reexen Technology, Graphcore and Mythic, etc. The top five players hold a share over 80%. North America is the largest market, has a share about 50%. In terms of product type, In-memory Computing is the largest segment, occupied for a share of about 88%, and in terms of application, Small Computing Power has a share about 90 percent.

Analysis of the market drivers of Processing-in-Memory (PIM) technology,

1. Explosive growth in computing power demand: the underlying pressure of AI and big data

Demand for AI training and reasoning:

The global AI chip market is expected to reach US$120 billion in 2025, of which 75% of computing power is consumed in data transfer (not computing itself).

Large-scale language models (such as GPT-5) have more than 10 trillion parameters, and processing-in-memory (PIM) can improve the efficiency of sparse matrix operations by 3-5 times.

Data center energy consumption crisis:

Global data center power consumption accounts for 1.5% of total power demand, and data transfer energy consumption accounts for 40% in traditional architectures. Processing-in-Memory (PIM) can reduce energy consumption by more than 50% by reducing the memory wall effect.

2. Moore's Law slows down: an inevitable choice for architectural innovation

Process bottleneck:

The cost of advanced processes (below 3nm) has soared, and the marginal benefits of increasing transistor density have diminished. Processing-in-Memory integrates computing units through 3D stacking processes (such as HBM3) to break through the limitations of planar processes.

Heterogeneous computing needs:

Scenarios such as AI and graphics processing require customized computing units. Storage and computing integration supports the collaborative design of the logic layer and the storage layer to improve the efficiency of dedicated accelerators.

3. New storage technologies mature: hardware foundation is ready

Non-volatile memory (NVM) rises:

New memories such as ReRAM, MRAM, and PCM have analog computing capabilities and are naturally adapted to the storage and computing integration architecture. For example, the resistance state of ReRAM can directly participate in matrix operations.

Storage-class memory (SCM) popularization:

SCM technologies such as Intel Optane and Samsung Z-NAND have been mass-produced, providing PIM with high-performance, low-latency storage media.

4. Edge computing and IoT scenarios: energy efficiency revolution

The computing power dilemma of end-side devices:

Devices such as autonomous driving, AR/VR need to process massive amounts of data locally (such as 8K video streams). Storage and computing integration can reduce power consumption by 70% and extend battery life by 2-3 times.

Real-time requirements:

Predictive maintenance in industrial IoT needs to respond within microseconds, and storage and computing integration reduces data processing latency from milliseconds to nanoseconds.

5. Software ecology and algorithm collaboration: application scenario expansion

Sparse algorithm optimization:

Sparse matrices account for more than 95% of neural networks, and storage and computing integration can skip zero-value calculations, improving efficiency by more than 10 times.

Programming model evolution:

PIM-oriented spatial computing paradigms (such as NDA and GenASM) are gradually maturing, and developers can call computing units in storage.

6. Policy and capital promotion: global technology competition upgrades

National strategic support:

The US CHIPS Act and the EU's European Processor Initiative both list storage and computing integration as key directions. China's "14th Five-Year Plan" clearly supports the development of storage and computing integrated chips.

Capital inflow:

In 2023, global PIM financing will exceed US$5 billion, and giants such as Samsung, SK Hynix, and TSMC will accelerate their layout, and start-ups such as Mythic and UPMEM will receive multiple rounds of financing.

7. Supply chain reconstruction: from vertical integration to open collaboration

Industry chain collaboration:

Memory manufacturers (Micron, Kioxia) and IP suppliers (Synopsys, Cadence) cooperate to develop PIM design tool chains.

Foundries (SMIC, UMC) launched 2.5D/3D packaging technology to support mass production of integrated storage and computing chips.

Summary: The integrated storage and computing technology market is driven by computing power demand, hardware innovation, and policy capital. The core competition will focus on process integration capabilities (such as 3D stacking), algorithm-hardware co-design, and ecological openness. Chinese companies need to overcome the shortcomings of memory media and EDA tools and accelerate the commercialization of AI and edge scenarios.

This report presents a comprehensive overview of the global Computing in Memory Technology market, covering market size and forecast, segmentation by product type and application, competitive landscape, leading players and regional and country-level outlook.

Segment by Type

• Near-Memory Computing
• In-memory Computing
• Processing In Memory

Segment by players, this report covers

• Syntiant
• Zhicun(Witmem) Technology
• Reexen Technology
• Graphcore
• Mythic
• Shanyi Semiconductor
• AistarTek
• Samsung
• SK Hynix
• Houmo Technology
• Pinxin Technology
• Yizhu Intelligent Technology
• TensorChip

Segment by Application

• Small Computing Power
• Big Computing Power

Who Can Use This Report?

This report is written for decision-makers who need a clear, data-backed view of the global Computing in Memory Technology market:

• Manufacturers, suppliers and solution providers benchmarking their position and planning product, capacity and go-to-market strategy
• Distributors, channel partners and end users in Small Computing Power, Big Computing Power evaluating demand and sourcing options
• Investors, financial analysts and consultants assessing growth opportunities, competitive dynamics and M&A potential
• Government agencies, industry associations and research institutions tracking industry developments and policy impact