Nvidia CEO speaks out under pressure

At Computex 2025, Nvidia introduced a series of new chip and software technologies. This is to meet the growing demand for AI computing, while maintaining the company's leading position in the market. The event attracted thousands of attendees in Taipei and hundreds of thousands of online viewers, despite the relatively dry information provided, which had little impact on end customers.

Appearing at the event, Jensen Huang still showed his confidence and attractive presentation style, mixed with a bit of humor. The 62-year-old businessman has become one of the most interested figures in the technology industry at the moment.

Nvidia's new products revolve around its chip ecosystem, a key component for developing and operating AI. The company's main goal is to expand AI's reach to many different industries and countries.

Nvidia’s event comes as the cost of AI training becomes a big question for many companies, while the tech industry awaits further action from the Trump administration on the global impact of tariffs.

Nvidia's position

As part of the main event of Computex 2025, the Nvidia Keynote seems to go beyond the tech expo. The American company held the show at the Taipei Music Center, about 4 kilometers away from Computex. The hall was larger, but the thousands of seats in the audience were packed with press, analysts, tech experts, and Nvidia partners from around the world .

Registration for the in-person event is limited, with the US company reviewing it days in advance. The semiconductor company is now worth $3 trillion , the second largest in the world. Every move it makes could have an impact on the entire tech industry in the current AI craze.

CEO Jensen Huang talks about the transformation of modern computing. Photo: Xuan Sang.

Jensen Huang opened the session by talking about how computing is changing around the world. The entire data center is becoming a computing unit. Modern computing is moving to these places, instead of individual devices or servers.

Mr. Huang said that Nvidia is no longer purely a technology company. They have become essential infrastructure for every other industry based on AI, something that no other company has done before.

The US company’s business opportunities have gradually changed. From the $300 million chip industry in 1993, they have moved to the $1 trillion data center opportunity. Right now, the company is on the verge of dominating the factory industry and AI infrastructure, estimated to be worth trillions of dollars.

New hardware technologies

On the product front, Jensen Huang started with information about the launch date of the GB300 Blackwell Ultra system, expected to be available in the third quarter of this year. The system is a significant performance upgrade compared to the Grace Blackwell, which is currently used by cloud service providers.

Huang also thanked the many suppliers, such as TSMC and Foxconn, that help build and distribute Nvidia’s technology around the world. “When you need to create new markets, it needs to happen here, at the center of the computing ecosystem,” Huang said of Taiwan, his home island.

Nvidia RTX Pro server system. Photo: Xuan Sang.

Next, Nvidia introduced the new RTX Pro Server system. For tasks related to DeepSeek, the RTX Pro Server's performance is 4 times higher than the previous H100 system.

The RTX Pro Server is also 1.7x more efficient on some tasks involving Meta's Llama model. Huang said the new product is in mass production.

Nvidia also introduced a new version of its complete line of computers for use in data centers. Specifically, NVLink Fusion technology allows customers to use Nvidia CPUs (with AI accelerators from other brands), or equip their own processors (with Nvidia AI chips).

Grace Blackwell chip and Nvidia's high-end AI server. Photo: Xuan Sang.

Previously, similar components could only use components provided by Nvidia. With the new technology, customers can choose their favorite CPU while still ensuring performance through the NVLink interconnect chip. The American semiconductor company's openness opens up opportunities for greater customization.

At the event, several companies including MediaTek, Marvell, and Alchip announced the development of AI chips for Nvidia's systems. Meanwhile, Qualcomm and Fujitsu also plan to develop custom processors to work with Nvidia's AI accelerators.

Software to support expansion and AI construction

Next, Nvidia announced that Acer and Gigabyte will join in customizing and distributing the company’s line of small-sized AI computers, including the Spark and Station. Currently, the list only includes Dell and HP.

To streamline the process of expanding chip factories, some companies like TSMC have used Nvidia's Omniverse platform software/services to create "digital twins" of factories.

Nvidia CEO introduces NVLink Fusion technology. Photo: TechPowerUp .

This process speeds up the upgrade and refinement of manufacturing facilities. This helps accelerate the expansion of partners’ AI factories, even supporting small, less experienced companies in building AI data centers.

In addition, Nvidia introduced new software that enables faster robot training through various custom scenarios. In addition, DGX Cloud Lepton software helps cloud service providers automate the process of connecting developers to computers, supporting the creation and running of services.

Source: https://znews.vn/ceo-nvidia-manh-mieng-truoc-suc-ep-post1554165.html