Nvidia To Start Mass-Producing China-focused AI Chips In Q2 2024: Report

Nvidia could begin mass production of an AI-based chip (the H20 chip) specifically designed for China in Q2 2024, according to a report by Reuters. The report cited two sources, one of whom said that the US tech company will produce a limited number of the chip at the start; the major customers most likely to be satiated first with the supply. Both sources reportedly “declined to be identified”.

Nvidia’s H20 Chip: “Most Powerful” of Its China-Focused Chips

The Nvidia AI chip we are talking about is the H20 chip. It is said to be the most powerful of the three microchips the US chipmaker is making for China, the other two being the L2 and L20.

Why Nvidia Is Making China-Specific Chips

Why does the US-based chipmaker has to produce China-specific chips? The US export regulations are the reason. Nividia Corporation got hit by the US restrictions when its flagship A800 and H800 AI chips were not allowed to be exported to China after new restrictions on the export of advanced chipsets and technology came in October.

Nvidia then modifed an advanced gaming chip in adherence with the US export laws and launched the new version in China in late December. Eventually, Nvidia started building China-focused chips to comply with the US export restrictions.

<img alt="Artificial Intelligence" data- data-src="https://kirelos.com/wp-content/uploads/2024/01/echo/1-1.png" data- data-wp-effect="effects.core.image.setButtonStyles" data-wp-effect–setstylesonresize="effects.core.image.setStylesOnResize" data-wp-init="effects.core.image.initOriginImage" data-wp-on–click="actions.core.image.showLightbox" data-wp-on–load="actions.core.image.handleLoad" decoding="async" height="400" src="data:image/svg xml,” width=”800″>

The chipmaker planned to launch the H20 chip in November 2023 but couldn’t and hence, the launch got delayed. The delay was reportedly caused due to issues faced by server manufacturers in integrating the AI chip. The three new AI chips are said to come equipped with Nvidia’s latest AI tech but to comply with the US rules, they’ll require less computing power.

Owing to the strict stance of the US government on exports, many Chinese firms are reportedly hesitant about going for the downgraded H20 in future and may have plans to choose China-based alternatives. For instance, Chinese search engine giant Baidu chose Huawei last year for its AI chips instead of US-based Nvidia Corporation.

Nvidia HGX H200 Tensor Core GPU

Nvidia recently launched its HGX H200 GPU, which, the company claims, is the “world’s most powerful GPU”. According to Nvidia, the HGX H200 IS “the first GPU with HBM3e, the H200’s larger and faster memory fuels the acceleration of generative AI and large language models (LLMs) while advancing scientific computing for HPC workloads.”

<img alt="Nvidia HGX H200 GPU" data- data-src="https://kirelos.com/wp-content/uploads/2024/01/echo/2-1.png" data- data-wp-effect="effects.core.image.setButtonStyles" data-wp-effect–setstylesonresize="effects.core.image.setStylesOnResize" data-wp-init="effects.core.image.initOriginImage" data-wp-on–click="actions.core.image.showLightbox" data-wp-on–load="actions.core.image.handleLoad" decoding="async" height="400" src="data:image/svg xml,” width=”800″>

The Nvidia HGX H200 GPU offers 141GB of HBM3e memory at 4.8 terabytes per second (TB/s). The company claims that what the HGX H200 GPU offers is “double the capacity of the NVIDIA H100 Tensor Core GPU with 1.4X more memory bandwidth.”

• Winner of The 21st Century Emily Dickinson Award for his debut book ‘The Heart of Heartbeats’, Utkarsh found his calling to be a writer while studying engineering and has been writing professionally for over six years now. After completing… read more