1

Cerebras Systems

Cerebras provides high-speed AI inference and advanced reasoning capabilities using its proprietary Wafer Scale Engine, enabling real-time processing and deployment for AI models.

+1

paid

United States

2

Groq

Groq offers fast AI inference solutions for openly-available models through seamless integration with platforms like OpenAI, emphasizing instant intelligence and quick model deployment.

-1

paid

United States

3

SambaNova Systems

SambaNova Systems delivers an enterprise-grade full stack AI platform to quickly deploy generative AI capabilities, offering superior performance and scalability for large-scale AI workloads.

paid

United States

4

Cambricon Technologies

Cambricon offers advanced AI acceleration cards and systems leveraging MLUarch03 architecture and chiplet technology for high-performance cloud and edge computing.

unknown

China

5

Tenstorrent

Blackhole is a next-generation chip designed to efficiently handle massive AI workloads, featuring an open-source stack and MLIR-based compiler for seamless integration with AI software.

unknown

Canada

6

Arm

Arm designs and licenses CPU, GPU, NPU, and compute subsystem IP for cloud and data center, autonomous machines, mobile and PC, and embedded IoT devices. It also provides developer tools, software, training, and a partner ecosystem for building Arm-based systems.

N/A

unknown

United Kingdom

7

Lightmatter

Lightmatter develops photonic chips and interconnects to advance AI infrastructure, enhancing processing capabilities and efficiency through silicon photonics innovation.

-1

unknown

United States

8

Graphcore

Graphcore offers cutting-edge IPU hardware and software solutions, enhancing AI model development and deployment across various domains, including finance, biotech, and telecommunications. It accelerates research and innovation through specialized machine intelligence solutions.

-1

paid

United Kingdom

9

Hailo

Hailo develops AI processors for edge devices, offering high performance in deep learning applications, generative AI, and video processing. These processors enhance performance and reduce power, ensuring efficient edge computations without compromising privacy.

-1

unknown

Israel

10

D-Matrix

D-Matrix introduces Corsair, a computing platform that provides efficient AI inference for data center scale. It focuses on redefining performance and efficiency for scalable AI solutions, addressing the limitations of existing models in commercial AI applications.

-1

unknown

United States

11

Rebellions

ATOM is a pioneering AI chipmaker delivering scalable and energy-efficient inference solutions with a focus on generative AI. It utilizes advanced semi-conductor engineering to cater to hyperscale and latency-critical workloads.

-1

unknown

South Korea

12

Axelera AI

Axelera AI delivers edge AI acceleration hardware and software that enables high-performance, energy-efficient inference for computer vision and LLM workloads. Its Metis AIPUs, Europa systems, and Voyager SDK help organizations deploy scalable AI solutions faster and at lower cost than GPU-based alternatives.

-1

paid

Netherlands

13

Mythic

The M1076 Mythic AMP™ delivers up to 25 TOPS in a single chip, enhancing high-end edge AI applications with power-efficient AI inference. Mythic’s technology eliminates memory bottlenecks by storing AI parameters directly in the processor, offering improved energy efficiency, cost, and performance.

-1

unknown

United States

14

T100 OPU

Neurophos develops an Optical Processing Unit (OPU), a photonic tensor-core chip that performs AI inference matrix multiplication in the optical domain on standard CMOS. It also targets data-center prefill inference workloads with high efficiency.

N/A

unknown

United States

15

RAIDEN

EdgeCortix develops fabless AI semiconductor hardware and software for edge and Physical AI inference, including the SAKURA-II AI accelerator, the RAIDEN chiplet platform, the MERA compiler and software framework, and the Dynamic Neural Accelerator architecture. Its products target embedded, enterprise, industrial, robotics, smart city, space, and aerospace/defense applications.

N/A

unknown

Japan

16

Kinara

Kinara Ara-2 processors enable high-performance AI applications at the edge, optimizing compute needs for computer vision and generative AI. They offer seamless integration with AI frameworks, ensuring low latency and high efficiency.

-2

unknown

United States

17

Etched.ai

Etched.ai builds frontier inference clusters with co-designed chips, racks, software, and manufacturing for best-in-class throughput, latency, cost, and power efficiency on prefill and decode workloads. It targets trillion-parameter MoEs, long context, and agentic workloads.

-4

18

Netra Smart Edge Studio

Netrasemi designs Edge AI System-on-Chips and chiplet-based architectures for on-device, real-time AI computing in surveillance, robotics, automotive, and industrial automation. Its product portfolio includes the NETRA R1000 MCU, C-series vision SoCs, and A-series AI SoCs, supported by software, system-on-modules, and development kits.

N/A

unknown

India

19

Taalas

Taalas develops a platform for turning any AI model into custom silicon, producing Hardcore Models that are more efficient than software counterparts. It also hosts large language models on silicon chips and offers associated software, an API, and a conversational chatbot interface.

N/A

unknown

Canada

20

ANAFLASH

ANAFLASH provides semiconductor solutions for edge AI, including the Reflex-CU near-sensor computing unit and Logic-EFLASH embedded flash memory IP. It also offers the Legato-Logic time-domain neural network processor IP for low-power, low-latency on-device AI.

N/A

unknown

United States

21

ARKA-GKT1

Azimuth AI builds software-defined custom SoCs that combine compute, mixed-signal integration, and AI acceleration for edge and embedded applications. Its first-generation platform-on-a-chip, ARKA-GKT1, targets smart utilities, industrial IoT, connected infrastructure, and intelligent sensing.

N/A

unknown

United States

22

BrainChip

BrainChip develops neuromorphic Edge AI processors for efficient, on-device processing of voice, vision, and sensor data. Its Akida chip supports always-on voice control, real-time audio and video analysis, gesture recognition, and sensor monitoring, enabling low-power AI across industries.

-7

unknown

United States

23

Neural Decision Processors

Syntiant provides a full-stack edge AI platform combining neural decision processors, MEMS microphones and sensors, and optimized machine learning models and tools. It also serves device makers in consumer electronics, smart home, healthcare, automotive, and industrial markets, with processors that run deep learning models on-device for always-on, ultra-low-power applications.

N/A

unknown

United States

24

Tensordyne Napier

Tensordyne designs AI inference chips and rack-level systems for data center workloads, built on a proprietary logarithmic-math architecture. Its flagship product, Tensordyne Napier, is an air-cooled inference system targeting hyperscalers, neo clouds, and enterprises running large mixture-of-experts and agentic models.

N/A

unknown

United States

25

AnalogML

Aspinity designs analog machine learning chips that perform always-on sensor inference in the analog domain, keeping digital processors asleep until a relevant event is detected, targeting sub-300 kHz sensor signals. It also develops a chip for gigahertz RF classification, serving counter-drone, industrial IoT, asset protection, and smart home security applications.

N/A

unknown

United States