近日AI人工智能芯片领域出现资本运作出现大乱斗现象，不过大家的目标都是直指英伟达。
根据 The Information 最新报道，英伟达竞敌 Groq 预计将在未来两周内完成由贝莱德领投的 3 亿美元融资，估值达到 22 亿美元。Groq 创始人是谷歌专用芯片 NPU 发明者之一 Jonathan Ross。今年 2 月，Groq 发布了演示视频，展示了其 AI 芯片如何在几分之一秒内运行 LLM。
而AMD超微半导体公司日前表示，将以约6.65亿美元现金收购芬兰人工智能初创公司Silo AI，超微半导体公司正试图增强其 AI 芯片能力，与行业领导者Nvidia竞争。
此次收购是 AMD为扩大其在人工智能领域的影响力而采取的一系列举措中的最新一步。2023年，AMD收购了人工智能软件公司Mipsology 和 Nod.ai，并在过去12个月内向十几家人工智能公司投资超过1.25 亿美元。
AMD表示，收购Silo AI将帮助 AMD 改进基于AMD的AI模型的开发和部署，并帮助潜在客户使用该公司的芯片构建复杂的 AI 模型。
另一芯片行业巨头ARM的控股企业日本软银集团近日也宣布，将以4亿英镑的价格收购英国人工智能初创公司Graphcore。Graphcore曾被誉为"英国英伟达"，是一家专注于生产智能处理器单元（IPU）的AI芯片制造商。
软银集团此次收购Graphcore，旨在进一步扩大其在半导体领域的版图，并实现与集团旗下英国Arm公司的协同效应。
这些收购，其实AMD对Silo AI的收购最令人关注。总部位于芬兰赫尔辛基的 Silo AI 专注于端到端 AI 驱动的解决方案，帮助客户将技术集成到他们的产品和服务中。这家初创公司在欧洲和北美开展业务，其客户包括飞利浦、劳斯莱斯和联合利华等公司。
而且AMD的掌门人苏姿丰（Lisa Su）小姐与英伟达的掌门人黄仁勋（Jensen Huang）先生俩人还是表兄妹关系，虽然俩人在现实生活中交往不多，但在商业圈里的交锋却火药味很重。
事实上，AI芯片领域各大科技公司都各有建树，在囤积英伟达AI GPU芯片的同时，也在增加自己的AI芯片部署与迭代。
如英特尔通过收购Habana Labs获得了Gaudi系列AI加速器的产品线；微软在2023年发布了两款自研的Azure Maia、Azure Cobalt 产品；亚马逊自研的Graviton AI芯片迭代到了Graviton4；meta（现称为meta Platforms）自研的MTIA（meta Training and Inference Accelerator，即meta训练与推理加速器）迭代到了MTIA v2；谷歌的AI芯片TPU（Tensor Processing Unit）系列迭代到了TPUv5p等。
不过目前AI芯片英伟达仍以其CUDA等软件开发平台和Tensor Core GPU结合优势，以及在OPEN AI应用中的绝对市场应用地位，占据了90%的AI芯片市场份额，暂时独霸AI芯片江湖！
Recently, there has been a big brawl in the field of AI artificial intelligence chips, but everyone's goal is to target NVIDIA.
According to a new report from The Information, Nvidia's rival Groq is expected to close a $300 million funding round led by BlackRock in the next two weeks, bringing the valuation to $2.2 billion. The founder of Groq is Jonathan Ross, one of the inventors of Google's dedicated chip NPU. In February, Groq released a demo video showing how its AI chips can run LLMs in a fraction of a second.
AMD Supermicro Semiconductor recently said it will acquire Finland artificial intelligence startup Silo AI for about $665 million in cash, and Supermicro Semiconductor is trying to enhance its AI chip capabilities to compete with industry leader Nvidia.
The acquisition is the latest in a series of initiatives AMD is taking to expand its presence in the field of artificial intelligence. In 2023, AMD acquired AI software companies Mipsology and Nod.ai and has invested more than $125 million in more than a dozen AI companies over the past 12 months.
AMD said the acquisition of Silo AI will help AMD improve the development and deployment of AMD-based AI models and help potential customers build complex AI models using the company's chips.
Japan's SoftBank Group, the holding company of another chip industry giant, ARM, also recently announced that it will acquire United Kingdom artificial intelligence startup Graphcore for 400 million pounds. Graphcore, once known as the "Nvidia of United Kingdom", is an AI chip manufacturer focused on producing intelligent processor units (IPUs).
SoftBank Group's acquisition of Graphcore aims to further expand its semiconductor footprint and achieve synergies with Arm in the United Kingdom.
Of these acquisitions, AMD's acquisition of Silo AI is the most noteworthy. Headquartered in Helsinki, Finland, Silo AI focuses on end-to-end AI-driven solutions that help customers integrate technology into their products and services. The startup operates in Europe and North America, and its customers include companies such as Philips, Rolls-Royce, and Unilever.
And AMD's head Miss Lisa Su (Lisa Su) and Nvidia's head Mr. Jensen Huang (Jensen Huang) are still cousins, although the two do not have much contact in real life, but the confrontation in the business circle is very gunpowder.
In fact, all major technology companies in the field of AI chips have their own achievements, and while hoarding NVIDIA AI GPU chips, they are also increasing their own AI chip deployment and iteration.
For example, Intel acquired the Gaudi series AI accelerator product line through the acquisition of Habana Labs; Microsoft released two self-developed Azure Maia and Azure Cobalt products in 2023; Amazon's self-developed Graviton AI chip has been upgraded to Graviton4; meta (now called meta Platforms)'s self-developed MTIA (meta Training and Inference Accelerator) has been upgraded to MTIA v2; Google's AI chip TPU (Tensor Processing Unit) series has been upgraded to TPUv5p, etc.
However, at present, NVIDIA still occupies 90% of the AI chip market share with its advantages in software development platforms such as CUDA and Tensor Core GPU, as well as its absolute market application position in OPEN AI applications, and temporarily dominates the AI chip market!
