The recent boom in artificial intelligence has sent Nvidia’s value soaring. The boost saw the company overtake Microsoft to become the world’s most valuable company in June this year, with a market value of $3.34 trillion.

Of course, to the company’s position is highly coveted by competitors, and one of the companies that aims to compete for territory in the chip sector is Samsung. The South Korean manufacturer, best known in Brazil for its cell phones and home appliances, invests millions of dollars in hardware entirely focused on artificial intelligence.

O Samsung’s main project to intensify this dispute is the System-on-a-Chip (SoC) Mach-1. The company announced the chip in March of this year and seeks to rival not only Nvidia, but also its biggest competitor in the HBM memory segment, SK Hynix.

Samsung’s main bet is the System-on-a-Chip Mach-1.Fonte:  GettyImages

Like Blackwell graphics chips, the Mach-1 has an architecture designed to accelerate tasks with artificial intelligence. The component is not intended to serve the end consumer, but rather to meet the needs of large companies and data centers.

What is Samsung Mach-1?

As an AI accelerator, the Mach-1 works to reduce the bottleneck between graphics processing units (GPUs) and high-bandwidth memories (HBMs). To do this, the chipset takes advantage of the company’s processors and low-power DRAM chips, thus maintaining high performance but with low power consumption — which, according to Samsung, is even lower than equivalent solutions from Nvidia.

One of the clients on the horizon is Naver Corporation, a South Korean internet conglomerate based in Seongnam, South Korea. According to a report by The Korea Economic DailySamsung will supply between 150,000 and 200,000 Mach-1 boards, with each unit priced at $3,800 — a deal worth nearly $752 million.

While the deal allows Naver to decrease its dependence on Nvidia’s supply, it also opens space for Samsung in the high-performance chip sector. Naver is expected to use Mach-1s in AI inference servers, that is, in the machine learning process.

Samsung's new SoC aims to deliver high performance by reducing bottlenecks between the GPU and memory.Samsung’s new SoC aims to deliver high performance by reducing bottlenecks between the GPU and memory.Fonte:  GettyImages

However, the Samsung intends to expand its operations to the North American market. Still according to the The Korea Economic Dailythe manufacturer would be in negotiations with Big Techs such as Microsoft and Meta.

Promise for 2025

Expectations are high, but the Mach-1 is still in development. The new AI acceleration chip is expected to enter mass production of prototypes later this year — plenty of time for Nvidia to further consolidate its presence in the sector.

The South Korean company’s intention is to become the main provider of hardware for artificial intelligence.

In a way, the Major AI service providers must be rooting for the fight. If Samsung’s introduction to the market is with a quality product, this will force Nvidia to launch equivalent solutions at a more competitive price.

Mass production of Mach-1 prototypes is yet to begin.Mass production of Mach-1 prototypes is yet to begin.Fonte:  GettyImages

In this competition, being headquartered in South Korea could be a huge advantage for Samsung. There, the company could gain highly influential clients looking to distance themselves from their dependence on Nvidia and, over time, expand their operations outside the country.

The process is slow and, of course, it may face other obstacles — including other competitors — over time. One of the possible rivals could be Microsoft, which will take advantage of Intel’s factories to produce its own chips — but without revealing which ones they will be.

Advances in research

Another million-dollar investment by Samsung is in laboratories dedicated to the development of new semiconductors. In March of this year, the South Korean firm opened new offices focused on chip construction for large-scale language models (LLMs) with a focus on inference — hence training and hosting these models.

According to the head of the futures business division Kyung Kye-Hyun, the Samsung plans to launch new chip designs and an “iterative model” capable of delivering better performance and meeting the growing demand for models, but without as much manufacturing or energy cost.

The fruits of this new sector of Samsung will still take some time to appear, but they reinforce the company’s interest in competing for territory with Nvidia.

Source: https://www.tecmundo.com.br/produto/286727-tentativa-samsung-impedir-supremacia-da-nvidia.htm



Leave a Reply

Your email address will not be published. Required fields are marked *