Recently, GPU leader Nvidia officially announced the completion of an agreement with the Japanese Software Bank, which will use cash and stock to invest 40 billion US dollars to acquire Softbank's chip manufacturer Arm. Arm will become a division of Nvidia, but the brand and headquarters location will remain unchanged. Nvidia will also set up an AI research center in Cambridge, UK to develop Arm-based supercomputers. Huang Renxun, the founder and CEO of Nvidia, emphasized that in the future, Arm's original patent licensing business model will be maintained and Nvidia's technology will be further added to Arm's ecosystem.
Japan's Softbank Group bought Arm for approximately US$30.9 billion in 2016 and became one of its subsidiaries. However, in July this year, it began to spread that Softbank would support the companies invested by its Vision Fund in order to raise funds. , In response to the impact of the epidemic, the news that it intends to sell the Arm subsidiary. Arm also started business cleanup at the same time, splitting its IoT service business group, IoT platform, and enterprise data management business, and transferred it to Softbank. He focused on the semiconductor IP business and was originally scheduled to complete the spin-off in September. Until September 13, Nvidia officially announced the purchase of Arm from SoftBank, and the transaction content did not include this IoT service business group that has been transferred to SoftBank. According to information disclosed by Nvidia, the transaction will include $21.5 billion in Nvidia stock and $12 billion in cash. However, SoftBank will not hold more than 10% of Nvidia. In addition, according to the additional consideration clause in the contract, SoftBank may also receive another $5 billion in cash or stock. Nvidia will also release $1.5 billion in stock to Arm employees.
In an investor conference call after the merger was announced, Huang Renxun personally pointed out three reasons why he must not buy Arm. The first reason mentioned is that he hopes to bring Nvidia's IP (intellectual property) into Arm. Ecological network, Arm has spent 30 years in the past and has created a huge IP ecosystem with a large number of partners and IP customers. “Unless the two companies merge into one, it is difficult for Nvidia's products to enter this ecosystem. "Z2z
Secondly, because Nvidia promised to start supporting the Arm data center ecosystem last year, and Huang Renxun’s goal is not only to support the Arm architecture CPU core or CPU chip, but to support the entire CPU platform, and it includes Nvidia’s GPU. , DPU (data processing chip), and even the software layer, system architecture, and AP architecture must provide first-level support for Arm to create a complete set of data center platform support. Arm has spent four years developing the data center ecosystem. Through mergers and acquisitions, Nvidia can directly own the human information center ecosystem. This is the key to Huang Renxun’s second purchase. The last reason for
, Huang Renxun sees the influence of AI at the edge after the two companies join forces. He believes that the future form of computing will be an autonomous computing system that uses AI extensively, which will also bring a lot of opportunities. For example, 5G edge computer rooms, automated machines of various sizes and forms, autonomous smart factories, etc., Nvidia has long spotted this area and has developed a lot of software that supports the Arm architecture. "Nvidia wants to create a cloud-to-end The future."
Huang Renxun simply summarizes the five benefits that the two companies can bring after the integration, creating a top company in the AI era, bringing Nvidia technology to the huge mobile and PC market through the Arm licensing model, and accelerating the Arm server CPU The development blueprint is to attack the 3 markets (data center, edge AI and IoT), expand the number of Nvidia computing platform developers from 2 million to 15 million, and the last item is the increase in the financial report such as gross profit margin and EPS.
In another global media conference call, Huang Renxun further explained that the edge application world he wants to target is that there are a large number of small devices that can continue to calculate automatically, which can provide various smart predictions nearby and provide faster Responsiveness, AI computing, network technology, and data processing technology will become more and more important. Whether it is cloud computing, HPC, edge data centers, robots, automatic computing equipment, IoT, regardless of size, computing architectures are very similar, "combining the two The company’s capabilities can bring AI to devices of all sizes."
Arm's commercial licensing model provides IP licensing to chip partners to collect licensing fees, and the other is the royalties for OEM or semiconductor companies producing Arm-related technology products. In the past four years, Arm's development strategy has made strides out of the mobile device field and invested heavily in four major areas, including data center chips, server CPUs, automotive technology architecture, and edge IoT devices.
In order to seize the IoT market, Arm launched a new authorization model Flexible Access last year, so that low-end chips do not have to pay authorization fees in advance. Customers can first see the details of various chip designs and select the desired IP products. Wait until the chip is officially produced before paying according to the number of chips shipped. This new authorization model covers nearly 75% of Arm's IP. After more than one year of launch, in addition to the original 30 customers, more than 60 partnerships have also added new licenses. The
ecosystem includes client device companies (such as Qualcomm, Samsung, Microsoft), infrastructure companies (such as AWS, VMware, Red Hat, Suse, Oracle, etc.), AI companies (such as Facebook), IoT and embedded devices Enterprises (such as Broadcom), automotive industry. Huang Renxun emphasized: “Arm can bring an ecosystem of more than 180 billion edge devices and thousands of partners to Nvidia, allowing Nvidia to integrate AI computing with the Arm ecosystem. This is a world that Nvidia was difficult to reach in the past.”
It is precisely because everyone knows that Nvidia is just looking at Arm's ecosystem, so they are particularly worried about whether Arm's business model will change after Nvidia's acquisition. In particular, Arm will become a division of Nvidia rather than an independent company. In media meetings, this is the most concerned topic. Simon Segars, CEO of
Arm, emphasized that Nvidia is very clear that Arm’s business model and neutrality are Arm’s most important values, so it will not change easily. Huang Renxun also declared that the two companies will jointly develop the technology blueprint, but will maintain the Arm open licensing model and business model. The only impact after Nvidia's purchase of Arm is, "In the Arm business model, adding Nvidia's technology and IP will only Will increase (Add), will not change the original practice. Maintaining this model is the key to ensuring Arm's success." He repeatedly emphasized.
However, the product lines of the two companies are not completely complementary. There are also duplicate product lines or technologies that compete with each other. One of them is that Nvidia uses the RISC-V open-source instruction set architecture, a CPU design technology that competes with Arm, and has been used for GPUs. The design of small embedded single-chip microcomputers, such as the Falcon single-chip microcomputer, which is useful for more than 15 GPU products, is designed with RISC-V. Including Tegra SoC and other GPU products. Nvidia needs to use about 300 million such chips every year, which are installed in the GPU. After
buys Arm, will Nvidia give up the use of RISC-V? Huang Renxun emphasized that no, this type of product already has users, and there are many places inside Nvidia that use RISC-V chips. Therefore, in the future, they will continue to invest and continue to use products, which means that Nvidia will continue to bet on these two products. A competing chip instruction set architecture.
Arm originally had a Mali product line of image processing chips for the automotive industry that has been developed for many years. Huang Renxun also maintains the current situation. Even if it is in a competitive relationship with Nvidia’s original product line, “As long as there are customers using Mali, they will continue to sell this type of product. Products." The
transaction has been approved by the board of directors of Nvidia, SoftBank and Arm, but it still needs to be approved by the UK, China, US and EU authorities. Nvidia estimates that it may take nearly 18 months to complete the acquisition. In order to win the consent of the British government, Nvidia made a number of announcements and investments, including a commitment to maintain SoftBank’s commitment to acquire Arm in 2016, maintain Arm’s company name and brand, and Arm’s intellectual property rights will continue to be registered in the United Kingdom.
The Arm headquarters will remain in Cambridge, UK, but will expand to establish a world-class AI research center. The main research and development areas include medical care, life sciences, robotics, autonomous vehicles, etc. In addition, NVidia also plans to launch a project to build an AI supercomputer for Arm CPUs. Wen⊙ Wang Hongren
