Nvidia to Resume Chip Sales in China - Latest Updates

Nvidia Resumes Applications for H20 AI Chip Sales to China

Nvidia revealed on Monday its intention to re-apply for licenses to sell its H20 artificial intelligence chips to China. This follows a period of fluctuating regulations, initially imposed by the Trump administration and subsequently reconsidered after a significant meeting.

Anticipated Approvals and New Chip Introduction

The company anticipates receiving the necessary approvals from the U.S. government shortly and plans to commence deliveries soon thereafter, as detailed in a recent blog post. Furthermore, Nvidia is launching a new RTX Pro chip tailored for the Chinese market.

This new chip is described as being “fully compliant” with current regulations and is specifically designed for applications within digital manufacturing, including smart factories and logistics operations.

The H20 Chip and the U.S.-China Tech Landscape

The H20 chip is central to the wider technological competition between the U.S. and China. Although not Nvidia’s most advanced AI processor, it represents the highest-performing chip the company is currently permitted to legally sell within China, adhering to existing export controls.

Its primary function is to facilitate “inference” – the process of utilizing pre-existing AI models for everyday tasks – rather than the development of new AI systems.

Stockpiling and Chip Appeal

Leading Chinese technology companies, including ByteDance, Alibaba, and Tencent, actively increased their stockpiles of these chips during the first quarter of the year. This was done in anticipation of potentially stricter export controls being implemented.

The chip’s attractiveness stems from its superior memory bandwidth when compared to Chinese-made alternatives, coupled with Nvidia’s established and widely-used software ecosystem, which simplifies hardware deployment.

Initial Restrictions and Subsequent Reversal

The initial restrictions on H20 sales were enacted in April by the Trump administration. These measures potentially threatened Nvidia with revenue losses estimated between $15 billion and $16 billion, based on the reported spending by Chinese firms in the first quarter.

The restrictions targeted chips exceeding specific performance criteria, including a total memory bandwidth of 1,400 gigabytes per second or an input/output bandwidth of 1,100 GB per second.

However, this decision was quickly reversed following a $1 million-per-head dinner attended by CEO Jensen Huang at Donald Trump’s Mar-a-Lago resort in early April.

Reports from NPR suggest the White House altered its stance after Nvidia pledged substantial investments in U.S. data centers. Subsequently, Nvidia announced plans to invest up to $500 billion over the next four years in building AI servers within the U.S., with support from partners like TSMC.

Criticism and Concerns Regarding China’s AI Advancement

This regulatory shift has faced criticism from U.S. lawmakers, who contend that it weakens the nation’s efforts to limit China’s advancements in artificial intelligence. They cite DeepSeek as an example of why these limitations are crucial.

The Chinese startup gained prominence earlier this year by developing a sophisticated AI model utilizing Nvidia’s H800 chips – a slightly more powerful predecessor to the H20. (Sales of the H800 chips were prohibited by the U.S. in October 2023, though Chinese suppliers have found ways to circumvent these restrictions.)

Ongoing Dialogue and Balancing Act

According to a statement provided to TechCrunch by Nvidia spokesperson Hector Marinez, CEO Jensen Huang has been engaged in discussions with officials in both Washington and Beijing this month.

He has been “emphasizing the benefits that AI will bring to business and society worldwide.”

This situation highlights the ongoing challenge faced by U.S. policymakers, who must balance national security concerns with significant commercial interests. Considering recent events, further reversals in policy are likely to occur.