黄仁勋柏林GTC力推欧洲全行业采用AI，发布DSX Pod并点名供应链扩产伙伴，但无新增订单或指引

Nvidia Corp. Chief Executive Officer Jensen Huang pushed Europe to adopt artificial intelligence more quickly, striking a bullish note as some investors grow more skeptical about the sums being poured into the infrastructure underpinning the technology.

The Santa Clara, Calif.-based company, whose chips have powered the artificial intelligence boom, unveiled new data center equipment, software, models and partnerships with leading companies on the continent at the event.

“For Europe to benefit from AI, and to have really created an AI ecosystem and AI economy, you need all the end markets, the end industries to adopt AI,” Huang said Wednesday at his company’s GTC conference in Berlin. He pointed to the automotive sector, financial services and health care. “All of these different industries have to become AI companies.”

The event is part of a tour through Europe by Huang, who has been signing deals with local companies and trying to persuade the continent’s political and business leaders that artificial intelligence is safe. His trip comes at a delicate moment for the industry. AI stocks have been battered during the past two weeks, and the closely watched Philadelphia Stock Exchange Semiconductor Index suffered its longest losing streak since 2008.

Read More: Nvidia CEO on AI Market Turmoil: ‘All of This Is Going to Pass’

Huang and his executive team hold GTC conferences around the globe, explaining the latest AI developments while aiming to drum up business for the company’s ever-expanding product lineup.

The new equipment includes the DSX Pod, a product that acts as a mini data center and can be used by companies trying to get hardware up and running more quickly. With one container, customers get the server hardware, cabling and plumbing they need — already installed and tested. It only needs to be plugged in and linked to other similar units, Nvidia said.

The new design, put together by Nvidia partners such as Dell Technologies Inc., will bring down the time required to install such hardware to 14 days from the current six months, Nvidia said.

The container is best suited to smaller-scale deployments of less than a gigawatt. Data center projects are typically measured by the amount of energy they consume since that’s a critical input.

The chipmaker said customers across Europe are adopting its DSX AI data center blueprint and newest processor design, Vera Rubin. Nscale Ltd., an Nvidia-backed data center developer, is among those companies.

Nvidia, the world’s most valuable publicly traded company, has frequently cited supply chain constraints as an impediment to its growth. To combat that, Ibiden Co., Samsung Electro-Mechanics Co. and Unimicron Technology Corp. will help it get more substrates — the materials used to house other components.

Amkor Technology Inc. will help increase the amount of chip packaging available by fourfold by the end of 2028. That material is used to encase semiconductors and connect them with other components. And KYEC Microelectronics Co. will increase testing capacity. Nvidia is also helping to boost the availability of on-package optical connections, which will speed up communication between components.

With concerns growing that AI will become too difficult for humans to control, Nvidia is developing software to ensure that so-called agents — programs that orchestrate the use of multiple models and programs — stay within prescribed limits. Nvidia is offering up a reference design combination of hardware and software that stops agents from accessing data they shouldn’t and prevents them from giving themselves permission to do things not explicitly allowed by their designers.

In so-called physical AI — a field that includes robots — Nvidia said it’s making rapid progress. Revenue from that market rose as much as 30% to about $11 billion in the first half of the year, the company said.

Nvidia announced a tie-up with a Siemens AG division for the deployment of AI-based advanced radiation therapy machinery. Such machines will monitor patient breathing and movements and adjust dosage dynamically.

The difficulty of deploying humanoid robots — one of the key mechanisms by which AI will become pervasive in the real world — remains a problem that many companies are working on using Nvidia’s chips and software.

The problems are so complex and expensive that the field is now only served by well-funded companies, Nvidia said.

To help open up the work to as many as possible, Nvidia is working with Agile Robots SE to create an open robotics platform that researchers can use to do development work. It’ll come with Nvidia’s Jetson Thor computer hardware inside of an Agile One SN humanoid.

In the automotive field, Nvidia said Volvo AB is making self-driving cars that use its chips. Volkswagen AG is using Nvidia products for commercial vehicles, and Bolt Technology OU is using the tech company’s software and gear for robotaxis.