On October 28, local time, after Nokia announced that NVIDIA will acquire shares in the company, NVIDIA CEOHuang RenxunAppeared at the GTC DC conference in Washington, USA, and also announced in his speech that Nvidia would invest in Nokia.

¡°Today we announce a partnership with Nokia, the world¡¯s second-largest telecommunications manufacturer, with a $3 trillion industry,¡± said Huang Yong, who said there are millions of base stations around the world and that the two sides will work together to transform the industry on the basis of new technologies, based on accelerated computing and AI.

This is the latest example of Nvidia's frequent large-scale investments in recent times. While investing and partnering with Nokia, Nvidia launched a new product line, Arc Aerial RAN Computer, to support 6G. During the speech, Huang Renxun also mentioned that the revenue forecast of GPU products "shows off muscles" and talked about other new products that Nvidia has deployed outside the GPU.

Nvidia CEO Huang Renxun delivered a speech at the GTC DC Conference

Another big investment

Nvidia has frequently made large foreign investments recently. Unlike the other two investments that transformed strong players in the AI field into close allies, Nvidia's US$1 billion investment in Nokia is mainly to break into the telecommunications field and launch its own new product line.

Nokia said that Nokia's board of directors has decided to issue 166,389,351 new shares in a private placement, allowing Nvidia to make a US$1 billion equity investment in Nokia. Provided that certain customary closing conditions are met, Nvidia will subscribe for these shares at a subscription price of US$6.01 per share. The deal will give Nvidia a 2.9% stake in Nokia. Nokia will adapt its 5G and 6G software to run on Nvidia chips, and Nvidia will explore ways to use Nokia's data center technology in its AI infrastructure.

Nvidia said in a subsequent press release that the investment will accelerate AI-RAN (Radio Access Network) innovation and promote the transition from 5G to 6G. Both parties will accelerate the development and deployment of next-generation AI-native mobile networks and AI network infrastructure. While announcing the investment, Nvidia also launched Arc Aerial RAN Computer, an accelerated computing platform that supports 6G.

¡°Nokia will use NVDIA-ARC as the future base station architecture, and NVIDIA-ARC is also compatible with the current Nokia base station architecture AirScale.We will use 6G and AI to upgrade millions of base stations around the world.This is the first time we can use AI technology to make wireless communication more efficient.¡±

In his speech, Huang said about cloud computing based on radio telecommunications networks. He said that the Internet made communication possible, while smart companies had previously built a cloud computing system on the Internet, and now NVIDIA is going to do the same thing on radio telecommunications networks, so that cloud computing can directly reach the edge scenarios that data centers cannot reach because there are base stations all over the world.

NVIDIA has announced that it will work with Nokia and T-Mobile U.S. as part of the 6G development to advance and test AI-RAN technology, with relevant trials to begin in 2026.

Following the announcement of investment and cooperation news, Nokia rose more than 26% and NVIDIA grew more than 5% in October 28th. NVIDIA share price exceeded $202 per share, market value reached $4.92 trillion, a record high.

Recently, Nvidia's frequent foreign investments have attracted much attention from the outside world. These investments are often accompanied by cooperation. Nvidia has also used this to promote product shipments and develop new product lines. In addition to investing in Nokia and launching related products, in September, Nvidia also announced a US$5 billion stake in CPU manufacturer Intel, and the two parties will cooperate to customize chips. Also in September, Nvidia announced that it would invest up to US$100 billion in large model vendor OpenAI. The cooperation includes Nvidia helping OpenAI deploy at least 10 GW of AI data centers that contain millions of Nvidia GPUs.

Revenue of two architectural chips will exceed $50 billion

In his speech, Huang also highlighted the progress of the chip business and quantum computing business.

Huang Renxun said that NVIDIA's Blackwell chips are in full production in Arizona, USA, and NVIDIA is still preparing for the production of chips using Rubin's next-generation architecture. Rubin may enter production in October next year or earlier."It can already be seen that this year and 2026 (natural year), Blackwell and Rubin can bring more than US$500 billion. This revenue will come from 20 million GPUs, and 6 million units have been shipped so far. The revenue of $500 billion will be five times the revenue generated by Hopper architecture chips from 2023 to 2025."

In his speech, Huang Yinglong highlighted the progress in the chip business and quantum computing business.

Huang Renxun said that cloud service providers are investing a lot of capital expenditures. Behind it, general-purpose computing platforms are transforming into accelerated computing platforms. Even without AI, this transformation is still going on.

In the field of quantum computing, Mr. Yuan announced that NVQLink can support computer control and calibration, quantum error correction, allow QPU (quantum processing unit) and GPU supercomputers to hybrid simulation, it is also fully scalable and can fit the quantum computers of the future. Mr. Yuan also announced that there are 17 different quantum computing companies supporting NVQLink.

Huang Renxun also said that Nvidia's CUDA can be regarded as designed for accelerated computing of GPU and CPU. Now CUDA-Q has been extended so that it can support GPU and QPU at the same time. Computing between GPU and QPU can be done in a few microseconds. "Move" in a time.

"Now we realize that it is crucial to connect a quantum computer directly to a GPU-driven supercomputer so that error correction can be performed, artificial intelligence calibration and control of the quantum computer can be performed, and running on the GPU and QPU. The right algorithm, the two computers working side by side, is the future of quantum computing." Huang Renxun said.

In the past, the chip industry accounted for about 5% to 10% or less of the trillion-dollar IT industry, because Excel, browsers and other applications do not need too much computing. But in the new era, a computer is needed to understand the context at all times, each step requires generating a lot of tokens (tokens), and now AI's "thinking" brings incredible computing load to the infrastructure.

"Moore's Law has basically ended. The question is what should we do? If we don't find a way to reduce costs?" Huang Renxun said that the AI industry has entered a cycle, that is, the more people use AI, the more profits AI can generate. More computers are put into use to make AI smarter. What the industry needs to do is to reduce costs. To reduce the cost of AI computing and maintain this virtual cycle, the answer is co-design.

"You can't just design a chip and hope everything's okay." Huang Renxun said that NVIDIA started from a blank sheet of paper and considered new computer architectures, chips, systems, models and applications. The company now has Spectrum-XGS, an Ethernet technology that can support cross-regional expansion, and can integrate data centers into gigawatt AI factories. By understanding AI models and spanning the entire stack, you can create future-oriented architectures and reduce token costs.

In the field of robotics, Yuan also introduced the company¡¯s new self-driving development platform DRIVE Hyperion, and said that the turning point of Robotaxi (self-driving taxis) will come, which will travel a trillion miles a year and produce 100 million cars a year in the industry.