About AI (Artificial Intelligence) Chip
The popular makers or manufacturers of AI chips include Google, NVIDIA, AMD, Xilinx, Graphcore, Intel, Wave computing, Via, Qualcomm, Thinci, Amazon, LG etc.
Artificial Intelligence is a field of computer science (CS). It is any technique which enables machines to mimic like humans to perform various tasks such as decision making with experiences, learning, problem solving, planning, speaking, reasoning, identifying patterns etc. Machine learning (ML) is subset of AI where as Deep Learning (DL) is subset of ML.
Now-a-days AI has made major advancement due to development of ML (Machine Learning) algorithms which can process wide range of data quickly and produce more accurate outputs. Moreover it requires very less data pre-processing by human developers/operators.
AI hardware architecture depends on computing requirements as per use cases. This relies on use of CPUs, GPUs, accelerators, FPGAs, ASICs, memory storage and so on. Different applications require different processing speed, performance features, hardware interfaces etc. Due to AI (Artificial Intelligence) market various segments such as compute, memory, storage and networking are likely to see higher demand.
There are numerous applications of AI which include healthcare, transport, finance, data security, surveillance, gaming, robotics, automotive, social media, entertainment, e-commerce, education, agriculture, astronomy, home automation, face recognition, speech recognition etc.
AI chip makers | Manufacturers of AI chips
Following table mentions list of AI chip makers.
| AI Hardware chip manufacturers | Description |
|---|---|
| AI CPU (Central Processing Unit) | • AMD develops hardware/software solutions for AI such as EPYC CPUs which can be used in deep learning (DL) and
machine learning (ML) applications. • Graphcore has developed Rackscale IPU (Intelligent Processing Unit) Pod for data centers. It is based on colossus processor developed by Graphcore. • Adapteva is offering AI chip called "Epiphany" which is 1024 core and 64 bit microprocessor. • Samsung has developed "Exynos" microprocessor which is designed for LTE products. It houses on-device & enhanced NPUs (Neural Processing Units). • IBM has developed AI chip called "TrueNorth" which is modeled as per human brain and houses about 5.4 billion transistors. • Xilinx has developed AI chipsets by name "Versal" or "Everest" which contain about 50 billion transistors. |
| AI GPU (Graphics Processing Unit) | • NVIDIA develops AI chipsets based on GPUs which include Tesla chipset, Xavier and Volta. • AMD offers Radeon Instinct GPUs for ML and DL applications. • Mythic AI offers GPU which delivers performance equivalent to desktop computer. • Imagination Technologies offers "PowerVR GPU" which is complete neural network accelerator for AI chips. It delivers 4 Tera operations per sec. |
| AI TPU (Tensor Processing Unit) | • Alphabet, parent company of Google has developed TPU ASIC. It is designed mainly for TensorFLow framework of Google. |
| AI FPGA (Field Programmable Gate Array) | • Intel offers FPGAs for AI applications. • Lattice semiconductor develops FPGA based machine learning solutions. • Xilinx VITIS AI development environment. |
| AI platform provider | • Wave computing offers Triton AI SoC (System on Chip) which helps AI developers to develop AI use cased using single AI platform. The company develops AI solutions which can be used in edge devices, servers, data centers etc. |
| AI accelerators | • NVIDIA |
| AI kits | • Company called Via develops "Edge AI Developer Kit" based on qualcomm processor and other parts/components. |
| Other AI chip makers | • Baidu has developed Cloud to Edge AI chip. • Qualcomm has developed "Cloud AI chip" whch can also be used in telecom products based on 5G standard. • Thinci is developing solutions for ML, DL, neural networks (NL) and Vision processing (VP). It is developing autonomous car and drones. • HiSilicon, a semiconductor unit of Huawei has developed AI chips. • LG has developed its AI chip called "LG Neural Engine". • SambaNova develops custom AI chips. • Kalway has developed chip for AI processing which can be used in edge devices and data centers. • Amazon is developing AI chip called AWS Inferentia which can be used for its own business. • Google cloud and NVIDIA has developed first AI on 5G platform used for various applications. |
AI hardware landscape is rapidly evolving with new makers and manufacturers. There are several prominent companies and organizations involved in manufacturing AI chips and its hardware specifically designed for artificial intelligence (AI) and machine learning (ML) tasks or operations.