Hwang In-hoon saw smart bodies as the core driving force for the next AI infrastructure expansion at the annual British Wida Shareholders Conference. He stated that AI had moved from “interest” to “useful” and that the logic of data centre construction had changed as the model began to generate income directly.

Data centre to production token

Wong In-hoon indicated that the traditional data centre was primarily responsible for the storage and transmission of documents, while the goal of the new generation of AI factories was to produce a continuous output token. For customers, the procurement of the British Wida system is not just about purchasing computing equipment, but also about building production facilities that generate direct income.

He described the AI industrial chain with a five-tiered structure, Energy, Chips and Systems, Infrastructure, Models, Applications, and thus emphasized that the coverage of Britain was beyond GPU. The company wants to position itself as a full platform supplier for the AI plant.

  • Almost 40 countries are deploying AI plants
  • International market-related revenues have exceeded $30 billion
  • The company reaffirms that it returns more than 50% of the free cash flow.

Vera CPU Aim at smart bodies

Huang In-hoon called Vera Rubin one of the most important products in company history. According to him, Vera CPU is not an extension of the generic CPU, but a new entrance designed around intelligent missions.

He explained that smart people had frequent calls to tools, access to databases, implement codes and repeat them. If the CPU response rate is insufficient, GPU will be idle and thus lower output efficiency of the AI plant. As a result, Inverda chose to design the CPU from scratch for low-delayed demand.

In-hoon added that Vera Rubin was not a single chip, but a set of platforms, including Rubin GPU, Vera CPU, NVLink, Spectrum-X, InfiniBand and Bluefield. The platform is now in a production phase and model manufacturers, public clouds and mega-clients are moving forward with deployment plans.

CUDA remains a software barrier

At the software level, Huang In-hoon described the CUDA X Coco ecology as the "crowd jewels" of Inweda. In his view, the unified architecture, the size of the developers and the number of applications created a cycle of continuous reinforcement, which was also the hardest part of the competition to replicate.

Huang In-hoon stated that CUDA currently supported over 7000 applications. CUDAX is also moving from the developers ' tool set to the toolbox for smarts as they become new applications. During the Shareholders ' Conference, Weeda also referred to BioNemo, a tool kit for digital biology and drug discovery.

Chinese operations still do not recognize revenue

When it comes to the Chinese market, Huang In-hoon is cautious. He revealed that the United States Government had granted a licence for H200 exports to Chinese clients, but that British Weida had not yet recognized the revenues and that there was uncertainty as to whether products could be successfully entered into the Chinese market.

At the shareholder question session, he also responded to the problem of chip smuggling. Wong In-hoon stated that the company ' s compliance system had intercepted suspected cases of smuggling on several occasions, and that the advanced AI data centre relied on complete hardware, software, network and follow-up support and that limited chips alone made it difficult to build sustainable systems.