Current artificial intelligence (AI) computing is mainly based on neural network algorithms represented by deep learning. Traditional CPUs and GPUs can be used to perform AI algorithm operations, but they are not designed and optimized for the characteristics of deep learning. Therefore, it cannot fully adapt to the characteristics of AI algorithms in terms of speed and performance.
The industry's demand for artificial intelligence chips that support parallel computing capabilities will be increasing. However, even Intel, which has many world-class engineers and a strong research background, will take three years to develop its own AI chips. For most companies, purchasing chips from these vendors or renting capacity from cloud GPU providers is one of the ways to develop powerful deep learning models. The AI chip suppliers introduced in this article may help companies choose suitable chips.
1. NVIDIA
Nvidia has been producing graphics processing units (GPUs) for the gaming world since the 1990s. Both PlayStation3 and Xbox use Nvidia graphics arrays. The company also produces artificial intelligence chips for Volta, Xavier and Tesla. Thanks to the generative AI craze, Nvidia has a strong year in 2023, reaching a trillion valuation and solidifying its position as the market leader in GPU and AI hardware.
NVIDIA's chipsets are designed to solve business problems across a variety of industries. For example, Xavier is the basis for autonomous driving solutions, while Volta is targeted at data centers. DGXA100 and H100 are Nvidia's successful flagship AI chips, designed for AI training and inference in data centers. Nvidia releases H200, B200 and GB200 chips; HGX servers such as the HGX H200 and HGX B200 that combine eight such chips; the NVL series and GB200 SuperPOD combine more chips into large clusters
Cloud GPU
For AI workloads on the cloud, Nvidia has an almost monopoly, and most cloud vendors only use Nvidia GPUs as cloud GPUs. Nvidia also launched its DGX Cloud product to provide cloud GPU infrastructure directly to enterprises
2.AMD
AMD is a chip manufacturer with CPU, GPU and AI accelerator products. For example, AMD's Alveo U50 data center accelerator card has 50 billion transistors. Accelerator can run 10 million embedded datasets and perform graph algorithms in milliseconds
AMD launched the MI300 responsible for AI training work in June 2023 and will compete with NVIDIA for market share in this market. As ChatGPT shows, the rise of generative AI and the rapid increase in demand have made Nvidia's AI hardware difficult to purchase, so startups, research institutions, enterprises, and technology giants have adopted AMD hardware in 2023.
AMD is also partnering with machine learning companies like Hugging Face to enable data scientists to use their hardware more efficiently
3. Intel
Intel is the largest manufacturer in the CPU market and has a long history of semiconductor development. In 2017, Intel became the world’s first AI chip company with sales exceeding the $1 billion mark.
Intel's Xeon CPUs are suitable for a variety of jobs, including data center processing, and have had an impact on the company's commercial success
Gaudi3 is Intel's latest AI acceleration processor. There is currently limited benchmarking of its performance since its public release in April 2024
4. Alphabet/Google Cloud Platform
Google Cloud TPU is a purpose-built machine learning accelerator chip that powers Google products such as Translate, Photos, Search, Assistant and Gmail. It is also available through Google Cloud. Google released TPU in 2016. The latest TPU is Trillium, the sixth generation TPU
Edge TPU is another accelerator chip from Google Alphabet that is smaller than a penny and designed for edge devices such as smartphones, tablets and IoT devices.
5. AWS
AWS makes Tranium chips for model training and Inferentia chips for inference. Even though AWS is the leader in the public cloud market, it starts building its own chips after Google
6. IBM
IBM will release its latest deep learning chip, the Artificial Intelligence Unit (AIU), in 2022. IBM is considering using these chips to power its Watson.x generative artificial intelligence platform
AIU is built on the "IBM Telum processor," which powers the AI processing capabilities of the IBM Z large server. Prominent use cases for Telum processors at launch include fraud detection
IBM has also demonstrated that merging compute and memory can lead to greater efficiency. These have been demonstrated in the NorthPole processor prototype
Apart from this, there are some startups that may have a place in the industry in not only the future
7. Alibaba Group Holding Limited
Alibaba Group Holding Limited produces inference chips such as Hanguang 800. The peak performance of "Hanguang 800" is 78563 IPS and the peak energy efficiency is 500 IPS/W. The computing power of one "Hanguang 800" is equal to 10 GPUs (graphics processors).
8. SambaNova system
SambaNova Systems was founded in 2017 with the goal of developing high-performance, high-precision hardware and software systems for high-volume generative AI work. The company developed the SN40L chip and raised more than $1.1 billion in funding
Notably, SambaNova Systems also leases its platform to enterprises. SambaNova Systems’ AI platform-as-a-service approach makes its systems easier to adopt and encourages hardware reuse for the circular economy
9. Cerebras Systems
Cerebras Systems was founded in 2015. In April 2021, the company announced the launch of a new AI chip model Cerebras WSE-2, which has 850,000 cores and 2.6 trillion transistors. There is no doubt that WSE-2 is a big improvement over WSE-1, which has 1.2 trillion transistors and 400,000 processing cores
10.Groq
Grop was founded by former Google employees. The company's LPU is a new model of artificial intelligence chip architecture designed to make it easier for companies to adopt their systems. The startup has raised approximately $350 million and produced its first models such as GroqChip processors, GroqCars accelerators, and more
The company focuses on LLM inference and released a benchmark of Llama-2 70B
The company said that in the first quarter of 2024, 70,000 developers signed up on its cloud platform and built 19,000 new applications
On March 1, 2022, Groq acquired Maxeler, which provides high-performance computing (HPC) solutions for financial services
