In the current AI development trend, Nvidia is already dominating the leading position in the AI chip market and actively expanding its influence to data centers and related areas. CEO Jensen Huang is attempting to build Nvidia into a one-stop service platform with all the key data center elements. It will no longer just be a high-end chip hardware supplier, but an "AI factory."
Nvidia hopes to use this strategy to widen the gap with its competitors. At the financial report meeting for the latest quarter, Huang stressed Nvidia's strong ability to design AI factories and claimed that Nvidia has all the necessary components to complete the AI factory's construction. Not only would revenue increase, but clients would become further reliant on Nvidia's products.
It is estimated that Nvidia has taken up over 80% market share for the AI chip market. Aside from GPU achieving huge success within the industry, Nvidia is also actively expanding software, data center design and service, and network technology. For example, programmers can fully utilize Nvidia's chip functions through CUDA, a proprietary API that has been deeply cultivated for 17 years.
In addition, Nvidia will be investing resources in InifiniBand high-speed networking. Five years ago, Mellanox Technologies, which was acquired by Nvidia, was focused on developing this technology. It can be assumed that InfiniBand has likely become a core technology for most AI training and deployment.
Aside from hardware and software, Nvidia has also been developing AI-optimized Ethernet that can be widely applied to traditional data centers. Chief financial officer Colette Kress states that it is estimated that the AI Ethernet business will generate several billion dollars of revenue within a year.
Additionally, Nvidia has been selling products including CPU and networking chips. These products can work together seamlessly within data center equipment and offer professional hardware and software solutions according to the needs of specific industries such as healthcare and robotics.
Competitors actively aiming to close gap, Nvidia facing risks
However, competitors are also actively creating layouts. AMD has recently agreed to acquire data center design and manufacturing company ZT Systems with US$5 billion, aiming to close the gap between it and Nvidia. AMD has made other acquisitions in recent years as well, including Xilinx and data center networking company, Pensando, in 2022.
Other chip suppliers including Intel, Cerebras Systems, and SambaNova Systems, are also offering services and systems that support the creation and operation of AI tools. Plug-and-play AI infrastructures are currently favored by many, which can help corporations deploy AI solutions quickly.
However, Nvidia's AI factory strategy is also facing some risks. With competition intensifying in AI computing and the construction trend slowing down, clients may search for other substitute plans. While Nvidia is in the process of expanding market influence, it could face issues with regulations, especially in Europe, where Nvidia's market behavior has already been investigated, even having its French offices raided.
Huang recently stated that Nvidia's role is mostly that of a coordinator and designer, the goal is to guide the construction of complicated AI infrastructures, but Nvidia will not be producing all the equipment. With Nvidia planning to release the next generation of the Blackwell AI chip by the end of 2024 or early 2025, Nvidia will be providing design service for server racks accommodating full computing equipment.
Huang stresses that Nvidia is dedicated to designing AI infrastructures that accommodate client needs and blend them into the ecosystem. He also points out that this strategy will consolidate Nvidia's leading position and push Nvidia toward long-term success in the AI area.
