the health strategist
multidisciplinary institute for continuous transformation
via health strategy and digital health
Joaquim Cardoso MSc
Chief Research and Strategy Officer (CRSO),
Editor in Chief and Senior Advisor
June 16, 2023
ONE PAGE SUMMARY
In the current state of the generative AI chip market, Nvidia’s has a dominant position and there struggles faced by its rivals, such as AMD and Intel.
AMD’s made a recent attempt to catch up with Nvidia by introducing a new AI chip, the MI300X, but Nvidia’s lead in the market remains significant.
- While AMD claims impressive performance for its new chip, it has yet to identify potential customers considering the product, and its focus on AI inferencing rather than training limits its competitive edge against Nvidia.
- Furthermore, AMD’s chip will not be widely available until the first half of next year, by which time Nvidia’s flagship H100 chip will have been in the market for 18 months.
Wall Street has identified other chip companies, such as Broadcom and Marvell, that could benefit from generative AI …
… but their sales are not expected to come from the market dominated by Nvidia.
Demand for processors used in training and applying large AI models is experiencing the greatest surge, resulting in increased stock market value for Nvidia.
- While companies like AMD and Intel struggle to compete with Nvidia in advanced AI chips, they anticipate an evolution in the generative AI market, driven by the use of smaller and more specialized models.
- Additionally, the need for data security and local processing may drive demand for CPUs and AI accelerators made by Intel.
However, chipmakers face challenges in anticipating the evolving demands of data centers, driven by the surge in services like ChatGPT.
- It is predicted that sales of CPUs may even decline as data center customers prioritize investments in AI accelerators.
There is a software challenge faced by Nvidia’s competitors.
- Nvidia’s dominance in the market is partly owed to the ease of programming its GPUs using Cuda software.
- While AMD is working with PyTorch, a widely used AI framework, it has a long way to go to match the software libraries and applications already developed for Cuda.
Despite the desire for a more diverse market, Nvidia’s current lead in the booming generative AI chip market remains unchallenged.
Nvidia’s dominance in the market is partly owed to the ease of programming its GPUs using Cuda software.
This is a One Page Summary of the article “Nvidia’s rivals struggle to gain ground in generative AI chip war”, published by the Financial Times, and authored by Richard Waters.
