Nvidia CEO Jensen Huang told the global press that its plan of setting up an AI R&D center and the second AI Supercomputer Center is in progress, and directly confirmed that Nvidia will start doing ASIC chip designs in a press conference on June 4.
Nvidia has played a dominant role in the global AI industry through its AI GPU advantage, and its role has even been upgraded from an e-sports enabler to the "Godfather of AI." This near monopoly has also driven many customers to invest in self-developed chipsets for ASICs, including Google, Microsoft, Meta, Amazon, and other cloud-based service providers (CSPs), to reduce the huge risk of over-dependence on Nvidia, who controls the GPU price and shipment. Broadcom, Marvell, and some Chinese IC design houses are also comprehensively scrambling to enter the market to capture the opportunities.
It has been rumored that Nvidia will set up a new department to enter the ASIC market. In response, Jensen Huang said "Yes!" for the first time to confirm the decision. Huang said he believes this will further expand the customer base. Although the CSP customers will become Nvidia's competitors, at the same time, all CSPs will still be Nvidia's customers through the ASICs services.
Huang explained that Nvidia has three major advantages that cloud customers can't do without. First, although Nvidia's AI chips are not cheap, the value is like a smartphone that replaces the camera, music player, and other devices to integrate all the features in one gadget, which is very cost-effective on the whole; secondly, Nvidia's CUDA has an extensive and rich ecosystem; third, customers invest in self-developed chips, which greatly increases the cost of the chips.
Johnny Hsiang-Lin Shen, General Manager of AIchip, an ASIC design house based in Taiwan, said that Broadcom is the leader in the ASIC market, and the threshold of competition in this field is constantly rising, with very high requirements for design, capital, and overall operation.
Shen sees Nvidia's decision to enter the ASIC market as a bad idea, which will conflict with its existing products. He took MediaTek as an example, that the company's ASIC business is mediocre, and its ASIC business gross margins are not high although customers are demanding. Shen said AIchip is not worried about the challenge from its rivals.
Regarding Nvidia's investment plans in Taiwan in the next five years, Huang said it will employ at least 1,000 engineers and is currently recruiting talent on an ongoing basis for the large-scale AI R&D Center. It is also planning to set up a second artificial supercomputer center, but the location has not yet been determined. Nvidia has received the Taiwan government's commitment of NT$6.7 billion subsidy for the AI R&D center and has promised to develop applications for the Omniverse platform with Taiwanese companies to produce industrial solutions for the global markets. Nvidia also plans to build computing platforms for the Taiwan AI R&D center and offer part of the computing power to Taiwan's academic and research institutions, partners, or new entrepreneurs to use in research and development.
Nvidia has completed the construction of the supercomputer "Taipei-1" in December 2023. From July 2024 to February 2027, 25% of the computing power of "Taipei-1" will be shared free of charge with Taiwan's academia and research and development community, including manufacturers, schools, research institutes, and start-up teams.
On the other hand, Hsinchu County Mayor Yang Wenke also pointed out that Nvidia has approached his staff and expressed interest in setting up a facility in Hsinchu County, which is located between the northern and central tech clusters of Taiwan, hinting that Nvidia's R&D center will be located in Hsinchu County.
