On January 8th local time, chip giant Nvidia announced several heavyweight gaming chip products and multiple collaborations with Chinese gaming companies and electric vehicle companies before the official opening of the CES conference. There are also reports that Nvidia is expected to ship three customized AI chips for the Chinese market in the second quarter of this year.
Multiple news boosted Nvidia's stock price that day. As of the close of the US stock market on January 8th, Nvidia's stock price surged by over 6.4%, with a market value close to $1.3 trillion. This also drove a sharp rise in US chip stocks, with AMD rising over 5% and Intel rising over 3% throughout the day. In the past year, driven by the global AI boom, Nvidia's stock price has risen by over 230%.
Nvidia did not respond to First Financial reporters regarding the shipment details of customized AI chips in China.
But it can be confirmed that Nvidia, which is in the midst of the artificial intelligence boom, is expanding its advantages in the consumer grade GPU field to accelerate the development of "local" artificial intelligence. The so-called "local" artificial intelligence refers to the ability to run AI programs on PCs or laptops used in homes or offices without the need to send information to the cloud.
Nvidia also announced that it will collaborate with game companies such as miHoYo, Tencent, and NetEase to develop AI digital human services. The NVIDIA ACE (Avatar Cloud Engine) microservice has made its debut on CES, a technology platform that uses generative AI to create virtual digital characters. It allows game, tool, and middleware developers to incorporate generative AI models into virtual digital characters in games and applications.
Gaming chips have always been Nvidia's flagship product, and in recent years, the company has begun to improve its gaming graphics cards, striving to create chips that support the operation of artificial intelligence models on PCs.
The release of the latest chip means that although Nvidia is the company most closely related to large servers and data center GPUs, it will also compete with Intel, AMD, Qualcomm, and others in the field of local artificial intelligence.
Nvidia has released three new graphics cards: RTX 4060 Super, RTX 4070Ti Super, and RTX 4080 Super, priced between $599 and $999. Nvidia has stated that it will provide these graphics cards for laptops from companies such as Dell and Lenovo.
In December 2023, Nvidia announced the launch of a modified version of its advanced gaming chip, GeForce RTX 4090D, for the Chinese market to comply with US export controls against China. The company stated that the RTX 4090D has achieved a "qualitative leap" in performance, efficiency, and artificial intelligence driven graphics, and will be available to Chinese customers starting in January 2024.
In response to the difference between the three newly released graphics cards and the 4090D, Nvidia told First Financial reporters, "In China's product lineup, the RTX 4080 Super will be placed under the RTX 4090D." The company also stated that the three new graphics cards will comply with export controls and can be sold in China.
According to public information, the price of the RTX 4090 D is 12999 RMB (approximately 1842 USD), which is equivalent to the price of the RTX 4090 product sold in the United States and far higher than the price of the RTX 4080 Super at 999 USD.
Nvidia stated that the new consumer grade GPU chips will mainly be used for gaming, but can still run artificial intelligence applications. For example, the RTX 4080 Super generates AI videos 150% faster than the previous generation model. Nvidia also stated that other software improvements recently announced by the company will increase the processing speed of large language models by five times.
Justin Walker, Senior Director of Product Management at NVIDIA, gave an interview to media outlets such as First Financial at CES's press conference. Walker stated that the shipment volume of RTX GPUs has reached 100 million, providing a huge installation foundation for PCs that support artificial intelligence applications.
He also stated that the new chip can be used on Adobe Photoshop's Firefly generator to generate images or remove backgrounds in video calls. Nvidia is still creating tools that allow game developers to integrate generative artificial intelligence into their games, such as generating conversations for non player characters.
The technology industry is looking for the best way to deploy generative artificial intelligence, which requires a lot of computing power. The cost of running artificial intelligence on the cloud may be incredibly high.
Microsoft and Nvidia's competitors have also introduced the "edge computing" solution. Instead of using powerful supercomputers via the Internet, devices are built with more powerful AI chips, and can run large language models or image generators.
Nvidia believes that applications can use cloud models to solve tricky problems, and use local artificial intelligence models to handle tasks that require quick completion. Both methods can solve different tasks in parallel.
"The GPU on the cloud can run very large language models and use all processing power to support very large AI models, while the RTX tensor core in the PC will run AI applications that are more sensitive to latency." Walker explained.