US lawmakers accuse NVIDIA of supporting DeepSeek – with possible consequences for national security. The company rejects this, while Washington allows new exports.
• US Congress raises allegations against NVIDIA’s collaboration with DeepSeek
• NVIDIA refers to civil use at the time of technical assistance
• Chip exports to China are once again in the political focus
Allegations from Congress: NVIDIA’s Role in DeepSeek
On January 29, 2026, a letter from Washington brought new unrest to the already tense debate about AI and export controls. In a letter to US Secretary of Commerce Howard Lutnick, Republican Representative John Moolenaar made serious allegations against NVIDIA.
The chairman of the House Select Committee on China said that the US chip manufacturer had helped the Chinese AI startup DeepSeek to technically optimize its models. These models were later used by the Chinese military. The accusation is based on internal NVIDIA documents that are available to the committee.
Moolenaar wrote that NVIDIA employees helped DeepSeek with “optimized co-design of algorithms, frameworks and hardware.” The aim of the help was to significantly increase the efficiency of training large language models.
The political context is particularly explosive: DeepSeek attracted attention at the beginning of 2025 with powerful AI models that used significantly less computing power than comparable US models. This raised concerns in Washington that China could catch up quickly despite US export restrictions.
Why NVIDIA’s help was considered unproblematic at the time
At the same time, the letter made it clear that the collaboration took place at a phase in which DeepSeek was officially considered a civilian actor. According to Reuters, NVIDIA treated the company as a “legitimate commercial partner entitled to standard technical support.” There were no indications of military use of the technology at the time. The letter essentially states that DeepSeek was neither listed on a US sanctions list nor was it publicly known to be military-related.
The documents evaluated by the committee refer to activities from 2024. According to this, DeepSeek-V3 is said to have only required around 2.8 million GPU hours for complete training with the H800 chip developed specifically for China. Moolenaar argued that this was well below what US developers would have to spend on so-called frontier models. It is precisely these efficiency gains that have undermined the effectiveness of US export controls, according to the accusation.
NVIDIA, on the other hand, emphasized that the Chinese armed forces are not dependent on US technology. In a statement, the company said it “makes no more sense for the Chinese military to rely on American technology than it does for the U.S. military to use Chinese technology,” reports Reuters.
H200 export permits intensify political debate
Just one day after the letter became known, another aspect came into focus. On January 30, Reuters reported that China had given DeepSeek in-principle approval to acquire NVIDIA’s more powerful H200 chips. However, the approval is linked to conditions that have yet to be worked out. The H200 are considered to be significantly more powerful than the previously used H800 models and are particularly politically sensitive.
The proximity is no coincidence. Only at the beginning of the month did the US government clear the way for the export of the H200 chips to China, but on the condition that they do not reach military end users. Critics in the US Congress see a structural problem here. In his letter, Moolenaar warned that deliveries to supposedly civilian customers in China would “inevitably” violate military end-use requirements. The DeepSeek case serves as an example of how difficult it is to separate civilian and military use in China’s technology ecosystem.
NVIDIA CEO Huang: “We support every developer”
On January 31st, NVIDIA boss Jensen Huang finally spoke publicly. He denied allegations of special treatment to Bloomberg. “Whenever developers want to use our software, we openly support them,” Huang said. Every AI developer in the world works with NVIDIA in some capacity, he added, and that’s exactly what the company is proud of. DeepSeek’s support was part of this open developer approach and not a political statement.
This means that two perspectives stand side by side without being reconciled. On the one hand, US lawmakers see technical assistance as a factor that may have indirectly strengthened China’s military AI capabilities. On the other hand, NVIDIA refers to standard industry practice and that no military use was foreseeable at the time of the collaboration. One thing is clear: the debate comes at a time in which the USA, on the one hand, wants to secure its technological leadership role, but on the other hand, continues to allow business with China. The DeepSeek case is therefore unlikely to remain the last chapter in the dispute over AI, chips and national security.
Editorial team finanzen.net
By the way: NVIDIA and other US stocks can even be traded on finanzen.net ZERO until 11 p.m. (without order fees, plus spreads). Open a depot now for free and secure a new customer bonus!
Selected leverage products on NVIDIA
With knock-outs, speculative investors can participate disproportionately in price movements. Simply select the lever you want and we will show you suitable open-end products on NVIDIA
The leverage must be between 2 and 20
Advertising