Media in the U.S.: China asks some tech companies to suspend purchasing NVIDIA H200 chips!

One sentence: China's move is steady, accurate, and tough! It has asked some technology companies to stop ordering NVIDIA H200 chips this week, and also urged companies not to stockpile American chips. It is currently studying whether to allow the import of such high-performance chips and how to import them—its core message is clear: we can no longer let the lifeline of high-end computing power be handed over to others.

Many people may still remember that the U.S. previously imposed wave after wave of restrictions on China's high-end chips, from A100 and H100 to various subsequent limitations, with the intention of choking China's AI development.

But unexpectedly, the Trump administration later made a concession, allowing NVIDIA to sell H200 to China, but with a harsh condition: 25% of sales revenue must be paid to the U.S. government.

In short, they have turned chips into a political tool, wanting to both make money from China and at the same time control our technological development.

China now stopping procurement and regulating imports is essentially not wanting to be led by such "conditional supply."

Previously, some tech companies thought "it's better to stockpile first," but upon careful consideration, how big is the risk of stockpiling American chips?

Their suppliers could cut off supply anytime, add restrictions anytime, let you buy today, and tomorrow issue a ban, leaving your chips either unusable or unable to keep up with technological iteration—pure waste of money.

More importantly, excessive reliance on imports would deprive domestic chips of the soil for growth—if everyone buys ready-made chips, who will invest in R&D?

The Chinese government's move aims to set the tone for the market: we can no longer blindly rely on imports; we need to create space and opportunities for domestic chips.

Speaking of which, we should mention the progress of our domestic chips in recent years. Take Shanghai's chip enterprise cluster, for example, the "four little dragons"—Muxi, Biren, Tiangong Zhi Xin, and Suyuan—have risen rapidly. In early 2026, Tiangong Zhi Xin was listed on the Hong Kong Stock Exchange, with an opening price increase of 31.54%, raising 3.7 billion HK dollars entirely invested in R&D.

This company started designing general-purpose GPUs in 2018 and has now iterated to Tian Gai Gen3, with cumulative shipments of 52,000 units, serving 290+ clients, and implementing 900+ applications in fields such as finance, healthcare, and transportation, with a gross profit margin reaching 50.1%.

There are also companies like Huawei Ascend and Cambric, whose product performance is gradually approaching international advanced levels. Huang Renxun himself couldn't help but feel anxious, warning U.S. companies that if they continue to relax, China will win the AI race.

More importantly, the entire domestic chip industry has now taken shape. In 2025, the revenue of Shanghai's integrated circuit industry exceeded 460 billion yuan, doubling in five years, gathering more than 1,200 enterprises, accounting for 40% of the country's industry talent and nearly 50% of innovation resources.

People used to think domestic chips were "not good enough," but now what? General-purpose GPUs have already achieved mass production, and AI chips can directly replace imported ones in many scenarios. Although the domesticization rate of the GPGPU market in China was only 3.6% in 2024, it is expected to reach 31% by 2029, and in the field of intelligent computing chips, it is even expected to reach a 60% market share.

What does this indicate? It shows that we now have the confidence to replace foreign chips—we are no longer in a phase where we cannot function without imported chips.

Looking at the U.S. sanctions, they have long become a "double-edged sword."

Previously, the U.S. frequently restricted chip exports, resulting in heavy losses for companies like NVIDIA, with inventory piling up and revenue declining. Huang Renxun himself admitted that if geopolitical tensions eased, the company could export $2-5 billion worth of chips to China every quarter.

Later, the U.S. had to adjust its strategy, allowing partial chip exports and even abandoning additional tariffs. Essentially, they couldn't bear it anymore—China is the world's largest chip market. Losing China means a difficult life for U.S. chip companies.

Yet, despite this, they still want to impose such "25% take" clauses, turning commercial activities into political transactions. This kind of behavior is unseemly, and who would want to indulge it?

Some might ask, will halting the purchase of H200 affect domestic AI research?

Short-term, some companies that rely on imported chips may face some pain, but in the long run, this is actually a good opportunity to drive industrial upgrading.

At the end of the day, the key is no longer "whether we can buy imported chips," but "whether we want to hand over the initiative."

The U.S. chip policy is capricious—allowing exports today, changing course tomorrow. Relying on others' policies for technological security is itself the biggest risk.

China's current actions aim to break this dependence, telling the world: we welcome fair and reasonable technology cooperation, but we will not accept hegemonic clauses attached with political conditions; we are willing to achieve win-win results with companies from all countries, but we must guard the bottom line of self-reliance and control.

And halting the purchase of H200 is just the beginning. In the future, we will achieve self-reliance and control in more core technology areas, firmly grasping the initiative of technological development in our own hands.

This is not only the way out for China's technology industry, but also an inevitable choice for a major country to become strong—core technologies cannot be bought or begged for, they can only be developed by ourselves.

Original: toutiao.com/article/1853715506674695/

Statement: This article represents the views of the author.