Is ASIC Chip Design a Threat or Opportunity for Nvidia?

Realizing the changing competitive landscape, Nvidia aims to enter the ASIC dedicated chip design new business.

According to reports from Reuters and Bloomberg in early February 2024, Nvidia is establishing a new business unit focused on designing ASIC dedicated chips for cloud computing, AI, and other fields. We proposed in our April 2023 report that the competition in the AI chip market is becoming increasingly intense, and Nvidia’s biggest competitors may be their largest customers, including cloud computing giants and other tech giants developing their own chips, such as Google’s TPU, Amazon’s Tranium and Inferentia, Microsoft’s Maia, Meta’s MTIA, and Tesla’s Dojo. We believe that Nvidia aims to enter the ASIC dedicated chip market to avoid losing market share to substitutes, or because it realizes the changing competitive landscape and that its greatest computing power moat may gradually diminish. With the evolution of chip design, ASIC dedicated chips can effectively enhance performance and reduce latency by increasing HBM memory; additionally, advancements in algorithms can also effectively reduce the demand for chip computing power. Currently, some AI model training, such as Tesla’s FSD, has adopted FP8 instead of the traditional FP16.

Is ASIC chip design a threat or opportunity for Nvidia?

We believe that ASIC dedicated chips, with lower ASPs, higher efficiency and lower power consumption in specific scenarios, as well as more flexible design cycles, will complement or replace general-purpose GPUs. Consulting firm 650 Group estimates that the global market size for data center ASIC chips will be $10 billion in 2024, doubling by 2025. Currently, Broadcom and Marvell are leaders in ASIC chip design, and other Taiwanese companies such as MediaTek, Wistron, and Creative are also actively participating. The market generally views Nvidia’s establishment of an ASIC chip design department positively, but we believe that considering the lower revenue potential (fixed usage, slightly lower prices) and profit margins of ASIC chips (Wistron’s FY23Q3 gross margin was 23.7%, while Nvidia’s data center gross margin for the same period was 72.81%), if downstream customers collaborate with them on R&D, it may replace some of the company’s general-purpose GPU sales, which could impact its revenue and profitability.

The rapid growth of the AI industry, where will tech giants go in the arms race for AI chips?

We believe that major tech companies are actively seeking alternatives to avoid excessive reliance on Nvidia. On one hand, they are purchasing a large number of Nvidia’s GPUs, and on the other hand, they hope to support AMD as a second supplier while also developing their own dedicated chips. Currently, Nvidia’s customers, including OpenAI, Microsoft, Google, and Meta, are competing to purchase the company’s AI chips (for example, Meta announced it will purchase 350,000 H100 chips in 2024, and Microsoft also announced it will significantly purchase B100 chips in 2024) to compete in the AIGC field. According to the same report from Reuters, Nvidia has discussed collaboration on ASIC chip design with tech giants such as Amazon, Meta, Microsoft, Google, and OpenAI. However, we also believe that some tech companies are currently collaborating with other dedicated chip design companies to develop their own chips to avoid dependence on Nvidia’s dominance. Beyond AI, Nvidia may also collaborate with companies in gaming, telecommunications, and automotive sectors to design chips.

OpenAI’s ambitious trillion-dollar “chip-making” plan, where opportunities and risks coexist for Nvidia.

According to reports from the Wall Street Journal and CNBC in early February 2024, OpenAI CEO Sam Altman aims to raise $5 to $7 trillion to enhance global chip manufacturing capacity and address limitations in business growth, including the burden of scarce and expensive chips. The amount being raised is already higher than the combined market value of Microsoft and Apple, while OpenAI’s current valuation is around $80 billion. If the plan can be initiated, chip companies, foundries, cloud vendors, and investors will participate together, driving leapfrog development in the AI industry. We believe Nvidia is likely to benefit from the capacity expansion brought by the plan and the enormous AI chip demand from OpenAI; conversely, Nvidia may also lose its leading position as tech giants “start anew,” thus facing both crises and opportunities.

Risk Warning

AITechnology implementation and advancement may not meet expectations. Since the application of ChatGPT has achieved certain success, major tech companies have accelerated and increased their efforts to layout the AIGC field, such as Meta establishing an AIGC team at the beginning of 2023, and Microsoft further integrating AI technology into its Azure and Bing services. As artificial intelligence is a high-tech field, it requires substantial initial R&D investment and time. The subsequent implementation of AI technology may be influenced by various factors, including corporate investment, macroeconomics, policies, and public opinion, leading to R&D progress not meeting expectations.

Intense industry competition. Currently, generative AI technology is still in the early stages of industry development, with single and multi-modal large models continuously being launched, empowering applications such as chat, search engines, and code editing. The industry has not yet formed a relatively stable competitive landscape, leading to intense competition. If market competition further intensifies in the future, some companies may fail to timely launch relevant products or their technology R&D may not meet expectations, potentially leading to market clearing due to fierce competition.

China-USCompetition intensifies.As the two leading countries in the field of artificial intelligence, many domestic companies in both China and the US are actively deploying AI-related technologies and products, promoting the landing of cutting-edge technologies such as AIGC and LLM. If competition between China and the US intensifies in the future, it may hinder the further promotion of AI industry-related applications.

The information related to individual stocks mentioned in this article comes from publicly available sources and does not represent a research coverage or recommendation of the relevant companies.