Nvidia Corp is racing to contain political fallout after a United States lawmaker accused the chip giant of helping Chinese AI firm DeepSeek build models later tied to military use. Chief Executive Officer Jensen Huang has responded by stressing that Nvidia’s platforms are open to “every developer,” a message that doubles as a defense of the company’s business model and a bet that global demand for its chips will outweigh mounting scrutiny.
The clash over DeepSeek has become a test case for how far a dominant US technology supplier can go in serving customers in China while Washington tightens export controls. It is also forcing Nvidia to explain what “supporting developers everywhere” really means when those developers operate in a system where the state has sweeping authority over data and advanced computing.
The lawmaker’s charge: Nvidia and a Chinese military pipeline
The immediate spark for the controversy was a detailed letter from the House Select Committee on the Strategic Competition between the United States and the Chinese Communist Party, which argued that Nvidia’s work with DeepSeek helped power artificial intelligence later used by China’s armed forces. According to that letter, NVIDIA technology development staff helped DeepSeek achieve “major training efficiency gains” on its models, support that went well beyond simply selling chips. The committee said NVIDIA provided this extensive support while treating DeepSeek as a civilian customer, even as China’s system requires all entities to cooperate with state intelligence.
In a separate section of the same correspondence, lawmakers argued that NVIDIA provided this, and that Even now, despite mounting evidence of Deep ties to the People’s Liberation Army, the company has not fully acknowledged the military dimension. That framing set the tone for subsequent reporting that described how Nvidia’s engineers allegedly worked closely with DeepSeek to optimize algorithms, software and hardware, a collaboration that critics say blurred the line between commercial assistance and strategic enablement.
Inside the “co-design” claims and training boost
At the heart of the dispute is how far Nvidia went in tailoring its technology to DeepSeek’s needs. One account said that Nvidia worked to “co-design” DeepSeek’s model, providing technical support that allowed the Chinese firm to extract near-frontier performance from “deprecated” H800 chips that were still allowed under US export rules. In effect, Nvidia’s guidance helped DeepSeek train a powerful system despite semiconductor constraints that were meant to slow China’s access to cutting edge compute.
Another description of the same relationship said Nvidia’s technical support and cut the time needed for full training. A separate report, citing internal records, echoed that “According to NVIDIA records, NVIDIA technology development personnel helped DeepSeek achieve major training efficiency gains,” a line that appeared in a summary of how According to NVIDIA the company’s help reduced the compute needed for DeepSeek’s full training. Those details have become central to the argument that Nvidia did more than simply ship hardware into a sensitive market.
Huang’s message: “every developer” and a global platform
Confronted with these allegations, Chief Executive Officer Jensen Huang has leaned into a simple, repeated message: Nvidia’s software and platforms are meant for everyone who builds on them. In a recent appearance, he said that software is used by developers everywhere and that on that basis, of course, every developer uses NVIDIA software. In another clip, he added that he is “very proud of that,” framing the company’s reach as a point of pride rather than a liability.
In a separate interview, Huang reiterated that Nvidia Corp supports “every developer” that uses its software, a line delivered after he was asked about claims that the company helped China’s DeepSeek improve its artificial intelligence model. A video segment summarizing his remarks described how Nvidia CEO Supports DeepSeek-related claims, underscoring that he is not backing away from the principle even as political pressure rises.
DeepSeek as customer, and Beijing’s gatekeeping role
While Washington debates Nvidia’s responsibilities, DeepSeek is moving to deepen its commercial ties with the chipmaker. One report said DeepSeek has turned into a buyer of Nvidia’s next generation H200 accelerators, noting that Although the U.S. earlier this month cleared exports of the chip to China, Beijing retains the final authority over whether shipments actually go ahead. That dual control structure means Nvidia must navigate both American export rules and Chinese approvals to serve a single customer.
Another account described how regulators in Exclusive China conditionally approved DeepSeek to buy Nvidia’s H200 chips, according to sources cited by WKZO’s Everything Kalamazoo, a station that broadcasts on 590 AM and 106.9 FM. That same summary said China’s decision came with conditions, reflecting Beijing’s own desire to manage how much cutting edge compute flows into private hands. A separate note on the deal emphasized that China and Beijing retain leverage over the timing and scale of any shipments, even after Washington signs off.
How Nvidia frames DeepSeek in its broader AI strategy
Huang has been talking about DeepSeek for more than a year, often as a way to illustrate how fast AI models are evolving and how much compute they will consume. In one earlier discussion, he referred to DeepSeek’s R1 model as “incredibly exciting,” according to a summary that quoted Jensen Huang praising the system’s reasoning capabilities. However, he also stressed that the market’s reaction to DeepSeek’s efficiency gains was overdone, arguing that the most compute intensive part of AI is still ahead.
In another set of remarks, Huang said that Reasoning models can consume 100 times more compute than today’s systems and that future reasoning models will consume much more compute, a view he used to argue that demand for Nvidia’s chips will remain strong for years. That logic helps explain why he is so adamant about supporting all developers, including those in China: the company sees a world where every major AI lab, from DeepSeek to US startups, will need vast amounts of hardware and software to stay competitive.
More from Morning Overview