‘Origin’ of pain: China struggles to redraw chip supply map amid Trump’s tariff mayhem

Ni Tao is IE’s columnist, giving exclusive insight into China’s technology and engineering ecosystem. His Inside China column explores the issues that shape discussions and understanding about Chinese innovation, providing fresh perspectives not found elsewhere.

Nvidia CEO Jensen Huang concluded his whirlwind 48-hour trip to China on April 19. During his visit, rumors swirled around Nvidia setting up a factory in the country.

Media reports suggested that Nvidia was considering establishing a joint venture in China to maintain its CUDA ecosystem and potentially preparing to spin off its operations in China.

Nvidia, however, dismissed the reports as completely false. Still, one message was clear: Huang’s appearance—trading his signature black leather jacket for a business suit—signaled that despite US export restrictions, the company isn’t walking away from China just yet.

Previously, Nvidia had developed the H20 chip specifically for the Chinese market. But in a fresh blow, the chip was recently added to the US export control list as Washington moves to limit China’s access to advanced AI hardware.

Caught in the crossfire

Despite having only 15-20 percent of the H100 chip’s computing power and about 80 percent of the inference performance of the A100, the H20 remains the most powerful Nvidia chip Chinese AI companies can legally get their hands on.

In 2024 alone, Nvidia generated $12–15 billion in China from H20 sales, accounting for 80 percent of its China revenue and over 90 percent of the domestic market for inference graphic processing units (GPUs) used in large AI models.

If anything, the meteoric rise of DeepSeek earlier this year shows that robust inference models might be trained without top-tier GPUs, sparking doubts over Nvidia’s domination fueled by an AI boom.

Still, demand for Nvidia GPUs and its CUDA software ecosystem in China has proven resilient. After a dip, the company’s stock rebounded steadily, underlining its strategic importance.

Since the start of 2024, however, market volatility has wiped out approximately $844 billion of Nvidia’s market value. With such high stakes, China is a market the company can’t afford to lose — which explains Huang’s second trip to the country this year.

Nvidia’s strategic maneuvers in China reflect the broader reality facing many American companies doing business in China. Caught in the crossfire of US-China trade tariffs and tech embargoes, many are struggling with growing uncertainty: torn between complying with US regulations and preserving lucrative business in China.

Unlike Nvidia, which suffers from direct chip export bans, many US semiconductor firms are reeling from China’s retaliatory tariffs.

New rules on ‘place of origin’

On April 11, the China Semiconductor Industry Association issued an urgent notice, citing updated Chinese customs rules for determining the origin of semiconductor products.

According to the General Administration of Customs, the place of origin for integrated circuits will now be determined by the location of the wafer fabrication—namely, the “tape-out” process—not by packaging and testing.

This means the “origin” of a chip, whether packaged or not, will now be reported as the foundry’s location.

Chip manufacturing traditionally involves three major stages: design, wafer fabrication, and packaging and testing. Under the new customs rules, if the fabrication process changes the first four digits of the HS code (e.g., from semiconductor materials HS 8541 to integrated circuits HS 8542), then the region of fabrication—such as South Korea—is considered the place of origin, regardless of where packaging occurs.

Previously, US chipmakers could exploit China’s lenient “place of origin” rules to sidestep tariffs by packaging their US-fabricated chips in places like Malaysia or Taiwan.

The new rule ties a chip’s “nationality” to its place of fabrication, highlighting the geopolitical weight of foundries in the global semiconductor supply chain.

That said, leading US chipmakers like Nvidia, Apple, AMD, Xilinx, Qualcomm, and Broadcom remain largely unaffected by these new tariffs, as they primarily operate as design houses, outsourcing chip fabrication to TSMC in Taiwan or Samsung in South Korea.

A crushing blow to IDMs?

Integrated Device Manufacturers (IDMs), which design and fabricate chips in-house, may bear the brunt of the new customs rules.

IDMs handle everything in-house—from chip design to fabrication to packaging and sales. Examples include Texas Instruments (TI), Analog Devices Inc (ADI), and Intel. TI produces over 80 percent of its wafers in the US, while about 50 percent of ADI’s fabrication is domestic.

The new origin rules threaten to significantly weaken American IDM firms’ competitiveness in China, driving their prices. Potential victims include central processing unit (CPU) giant Intel, analog chip makers like ADI and TI, and memory maker Micron.

Beijing’s policy shift aims to kill two birds with one stone. It is poised to disrupt Washington’s push to restore chipmaking while supporting China’s strategy of strengthening its mature node foundry capabilities and building tech self-sufficiency.

For years, China has embarked on a “domestic substitution” campaign across critical sectors—from advanced chips to high-end industrial equipment, from narrow-body passenger jets to new energy supply chains—intent on reducing reliance on the West, especially the US.

Quiet reversal

Yet just two weeks after the new origin rules were issued, China quietly backpedaled, granting tariff exemptions to several US chip companies in a major, unilateral step away from its retaliatory stance.

On April 24, a document outlining China’s latest US tariff exemption list circulated online. It included eight categories covering semiconductor chips and equipment such as lithography, etching, and ion implantation tools. Notably, memory chips were excluded from the exemptions.

According to Caijing magazine, a business and financial news outlet, a Shanghai-based analog chip company, officials notified them on April 24 that some imports from US semiconductor firms would be exempt from tariffs, with retroactive rebates available for tariffs paid between April 10 and 24. They will only be subject to a 13 percent value-added tax.

This dramatic policy reversal illustrates one thing: despite China’s public claims of self-reliance in 28nm mature process nodes, many downstream users still rely on US chips to operate—and sometimes to stay afloat. One hundred percent self-sufficiency remains a long way off. The chip chokehold remains.

China’s analog chip localization rate rose from 9 percent in 2019 to 16 percent in 2024, but the proportion is still deemed too low in the ongoing “tech self-sufficiency” drive.

US giants TI and ADI continue to dominate the market. In 2024, TI’s analog business generated $12.2 billion, or 78 percent of its total revenue, with about $3 billion from China. ADI’s entire $9.5 billion revenue also came from analog chips, with 23 percent, roughly $2.2 billion, sourced from China.

US companies are estimated to generate $7–8 billion from China’s analog chip market in 2024.

The limits of ‘domestic substitution’

According to the China Business Industry Research Institute, a market think tank, China’s analog chip market reached RMB 302.6 billion ($41.6 billion) in 2023, up 9.05 percent year-on-year. The market is projected to hit RMB 325 billion ($44.7 billion) in 2024 and RMB 343.1 billion ($47.2 billion) in 2025.

Chinese customs data shows that China imported $11.8 billion of integrated circuits and $4.5 billion of semiconductor manufacturing equipment and components from the US in 2024.

While the domestic market’s size creates a major opportunity for Chinese firms to localize analog chip production, the reality is that, from electronic design automation (EDA) tools to wafer fabrication and packaging, Chinese companies still lack the capabilities and capacity to replace US suppliers at scale.

There’s also the issue of downstream demand. Even if Chinese firms could fill the gap, their customers—many of whom make export-oriented electronics, industrial gear, and car parts—face falling overseas sales due to US tariffs, which indirectly hit chip demand.

China exported only 116,000 cars to the US in 2024, a mere two percent of its total auto exports. This has so far insulated the Chinese car industry from the tariff fallout. Unlike electric vehicles, where US tariffs have limited impact, chips remain China’s biggest underbelly in the tech and trade war.

In an era when chips are essential to all electronics and constitute critical infrastructure for AI, the road to self-reliance remains difficult. Trying to force US firms to pick China by relocating foundry work won’t work. Harsh reality demands compromise.

That’s why Huang’s visit and subsequent rumors of Nvidia building a factory in China caused such a stir. Despite jokes about how DeepSeek’s rise may have cost Huang a few leather jackets’ worth of wealth, many in China hope those rumors are true.