DeepSeek, a Chinese AI startup founded in 2023, may have circumvented US chip export restrictions by using thousands of advanced processors dismantled in third countries and smuggled into China to train its latest V4 model, according to reports from technology outlet The Information. The allegations raise critical questions about how the company achieved its claimed cost advantages while operating under strict US sanctions on semiconductor exports to China.

How Did DeepSeek Obtain Advanced Chips to Train Its Latest AI Model?

In December, technology outlet The Information reported that DeepSeek developed its V4 model using thousands of chips that had been dismantled in third countries and smuggled to China, citing six people with knowledge of the matter. This alleged workaround would allow the company to bypass a strict US ban on exporting top-end chips directly to China, which has been in place to limit China's AI capabilities.

The allegations matter because they directly undermine DeepSeek's public narrative about achieving advanced AI performance at a fraction of typical costs. If the company relied on smuggled hardware rather than developing alternative solutions, the claimed efficiency gains become harder to verify and potentially misleading to investors and competitors analyzing the company's competitive advantages.

When asked about the smuggling allegations, DeepSeek did not respond to requests for comment. Nvidia, the US chipmaker whose processors were allegedly involved, also declined to comment but told The Information that they had not seen evidence of such smuggling and that "such smuggling seems farfetched". The company's dismissal of the allegations, however, does not resolve the underlying questions about how DeepSeek obtained the computing power necessary to train its models.

What Hardware Is DeepSeek Actually Using Now?

In response to the smuggling allegations, DeepSeek announced in December that it had used both Nvidia and Huawei's domestically-produced Ascend chips for its V4 model. This public statement appears designed to address concerns about sanctions violations by pivoting to legally available alternatives. Huawei, which leads Chinese efforts to reduce reliance on US semiconductor technology, confirmed in a statement that the full range of its Ascend products support DeepSeek's V4 series.

The shift to Huawei chips represents a significant development in China's broader strategy to build indigenous semiconductor capabilities that can operate independently of US technology. By publicly endorsing DeepSeek's use of Ascend processors, Huawei signals confidence in its own hardware while helping DeepSeek deflect concerns about sanctions violations.

Steps to Understanding DeepSeek's Supply Chain and Competitive Position

• Smuggling Allegations: The Information reported in December that DeepSeek may have used thousands of Nvidia chips dismantled in third countries and smuggled to China to train its V4 model, citing six sources with knowledge of the matter.
• Company Response: DeepSeek stated it used both Nvidia and Huawei's domestically-produced Ascend chips for V4, with Huawei publicly confirming that its full product range supports the model.
• Industry Skepticism: Nvidia stated it had not seen evidence of smuggling and called the scenario "farfetched," though the company did not provide detailed information about how DeepSeek obtained its hardware.
• Geopolitical Context: The allegations occur within the broader US strategy to restrict advanced chip exports to China, making hardware sourcing a critical vulnerability for Chinese AI companies.

Why Does DeepSeek's Hardware Strategy Matter for Global AI Competition?

DeepSeek burst onto the global stage in January 2025 with the release of its R1 deep-reasoning large language model, a type of AI system trained on massive amounts of text data to understand and generate human language. The release sparked a significant sell-off in US technology stocks because industry insiders were stunned by R1's performance, which matched leading US chatbots like ChatGPT, combined with DeepSeek's claims to have developed it at a fraction of the typical cost.

Venture capitalist Marc Andreessen called it a "Sputnik moment," referencing the 1957 launch of Earth's first artificial satellite by the Soviet Union, which shocked the Western world and triggered a technological arms race. That comparison underscores how seriously Silicon Valley views DeepSeek's emergence as a credible competitor. However, if the company achieved its cost advantages through smuggled hardware rather than genuine technological innovation, the competitive threat becomes more about sanctions evasion than engineering superiority.

DeepSeek currently holds four percent of global market share for chatbots based on web traffic analysis from Similarweb, while ChatGPT dominates at 68 percent. Despite the smaller market share, the trajectory matters more than current numbers; DeepSeek achieved this position in roughly two years, suggesting rapid growth potential if it can maintain reliable access to advanced computing hardware.

The company's success has energized China's entire AI sector. Shares in two leading Chinese AI startups, Zhipu AI and MiniMax, soared on their market debuts in Hong Kong this year, and Chinese chipmakers such as MetaX experienced similar gains. Investors at Beijing-based Jinqiu Capital observed a "clear surge" in enthusiasm around Chinese AI and increased competition among investors since DeepSeek's breakthrough.

What Does DeepSeek's Open-Source Strategy Add to the Chip Controversy?

DeepSeek's systems are open-source, meaning their inner workings are public and programmers can customize parts of the software to suit their needs. This approach contrasts sharply with the "closed" models sold by OpenAI and other Western rivals, which function as proprietary black boxes to end users. The Chinese government has actively promoted this open-source philosophy as a competitive advantage, with officials arguing that transparency accelerates innovation across the entire industry.

However, the open-source nature of DeepSeek's models does not resolve questions about the hardware used to train them. Even if the code is publicly available, the underlying computing infrastructure remains proprietary and subject to international trade restrictions. This creates a paradox: DeepSeek can share its software freely while potentially relying on illegally obtained hardware to create that software in the first place.

Like other Chinese chatbots, DeepSeek's AI tools avoid topics typically censored in China, such as the 1989 Tiananmen crackdown. This content moderation approach, combined with data privacy concerns and now questions about hardware sourcing, has led DeepSeek AI to be banned or restricted on government-issued devices in several countries, including the United States, Australia, and South Korea.

The tension between DeepSeek's technological achievements and geopolitical concerns highlights a broader challenge: how the world manages competition in artificial intelligence when national security interests, trade restrictions, data privacy, and innovation incentives all intersect. For now, the unresolved questions about the company's hardware sourcing remain a critical vulnerability that could undermine its credibility if evidence of sanctions violations emerges.