# Nvidia's Jensen Huang Defends Chinese AI Models, Opposes US Ban Push > Source: > Published: 2026-07-23 11:22:54.801041+00:00 # Nvidia's Jensen Huang Defends Chinese AI Models, Opposes US Ban Push - Jensen Huang told Axios that Chinese open-source AI models are "excellent" and American companies should "absolutely" be allowed to use them [[1]](https://www.axios.com/2026/07/22/nvidia-jensen-huang-china-open-source-ai) - Huang dismissed security fears about backdoors as a "misconception" and argued free AI models drive demand for Nvidia's chips and data centers [[1]](https://www.axios.com/2026/07/22/nvidia-jensen-huang-china-open-source-ai) - His comments directly contradict Treasury Secretary Scott Bessent, who threatened to sanction overseas AI developers found stealing American IP [[3]](https://stocktwits.com/news-articles/markets/equity/nvda-jensen-huang-american-companies-absolutely-use-chinese-ai-models-bessent-sanctions-threat/cZZm0d1R7wU) - The interview comes days after Moonshot AI released Kimi K3, a 2.8-trillion-parameter model that rivals top US models on key benchmarks [[4]](https://www.tomshardware.com/tech-industry/artificial-intelligence/moonshot-releases-2-8-trillion-parameter-kimi-k3) - Bipartisan legislation including the "No Adversarial AI Act" is advancing in Congress to ban Chinese AI models from federal agencies [[5]](https://www.tomshardware.com/tech-industry/artificial-intelligence/trump-administration-reportedly-reviving-push-to-ban-chinese-ai-models-following-kimi-k3-launch-citing-cybersecurity-concerns-downloadable-open-weights-could-make-an-outright-u-s-ban-nearly-impossible-to-enforce-amid-growing-adoption) Nvidia CEO Jensen Huang said American companies should be free to use Chinese open-source AI models and warned against a growing campaign to ban them, putting the head of the world's most valuable chipmaker directly at odds with the Trump administration and key lawmakers on both sides of the aisle. "These Chinese models are excellent," Huang told Axios in an exclusive interview published Tuesday. "Open-source models that are excellent should be used" [1]. Huang's remarks amount to the strongest public defense of Chinese AI from a major US tech executive at a moment when Washington is moving in the opposite direction. Treasury Secretary Scott Bessent has threatened to sanction overseas AI developers found stealing American intellectual property, telling reporters that officials discovered watermarks from American-built models embedded in Chinese AI systems [3]. Bipartisan legislation in Congress, including the "No Adversarial AI Act," would permanently bar federal agencies from using AI models developed in China, Russia, Iran, or North Korea . [[5]](https://www.tomshardware.com/tech-industry/artificial-intelligence/trump-administration-reportedly-reviving-push-to-ban-chinese-ai-models-following-kimi-k3-launch-citing-cybersecurity-concerns-downloadable-open-weights-could-make-an-outright-u-s-ban-nearly-impossible-to-enforce-amid-growing-adoption)The interview landed less than a week after Beijing-based Moonshot AI released Kimi K3, a 2.8-trillion-parameter mixture-of-experts model that ranked third on Artificial Analysis's Intelligence Index — behind only Anthropic's Claude Fable 5 and OpenAI's GPT-5.6 Sol — and first in blind developer testing for frontend code generation [4]. Nvidia shares closed at $212.06 on July 22, up 2.3% on the session, before slipping roughly 1% in pre-market trading Wednesday . [[6]](https://www.marketbeat.com/instant-alerts/nvidia-nasdaqnvda-shares-up-23-time-to-buy-2026-07-22/)## What Huang Said In the Axios "Behind the Curtain" video interview with Mike Allen, Huang argued that restricting access to open-source AI models — regardless of origin — would weaken the US technology ecosystem rather than protect it. "There's no scenario where China runs U.S. companies off the road. Zero possibility," he said [3]. Huang dismissed concerns that Chinese models could contain hidden backdoors or surveillance capabilities, calling such fears a "misconception" and noting that downloadable open-weight models allow users to customize their own guardrails [2]. On accusations that Chinese labs used distillation to copy American models, he was equally blunt: "Distillation, learning from AI, learning from other sources of knowledge, is fundamental to intelligence" . [[1]](https://www.axios.com/2026/07/22/nvidia-jensen-huang-china-open-source-ai)He framed his argument partly in commercial terms: open-source AI models, by lowering the cost of inference, expand the universe of companies adopting AI — which in turn drives demand for the GPUs and data-center infrastructure Nvidia sells. "Free AI should be great for hardware. Free AI should be great for chips. Free AI should be great for data centers," he said [3]. ## The Kimi K3 Catalyst Huang's comments are inseparable from the July 16 release of Kimi K3 by Moonshot AI. The model — with 2.8 trillion parameters, native vision capabilities, and a one-million-token context window — scored an overall Elo of 1,547 on Artificial Analysis's private evaluation, a 732-point jump from its predecessor Kimi K2.6 [4]. It beat Anthropic's Claude Fable 5 in the Frontend Code Arena benchmark in blind developer testing, while its API pricing of $15 per million output tokens undercuts Fable 5's $50 . [[2]](https://fortune.com/2026/07/22/jensen-huang-chinese-open-source-ai-models-kimi-deepseek-washington-ban-nvidia-chips-data-centers-security/)Moonshot has promised to release K3's open weights by July 27, which would make it the largest open-weight model ever released [4]. Huang told Axios the market had "misunderstood the impact of DeepSeek the first time" in January 2025 and has "misunderstood the impact of Kimi again this time" — implying that both events ultimately benefit the broader AI infrastructure buildout . [[1]](https://www.axios.com/2026/07/22/nvidia-jensen-huang-china-open-source-ai)## Washington's Counter-Push Huang's defense of Chinese AI runs headlong into a bipartisan consensus forming in Congress. The "No Adversarial AI Act," introduced by Rep. John Moolenaar (R-MI) and Rep. Raja Krishnamoorthi (D-IL), would permanently bar all US federal agencies from using AI models developed by entities in China, Russia, Iran, or North Korea [5]. Separately, a bill advanced by NBC News would target Chinese AI companies accused of copying American technology, while the Foreign Adversary AI Risk Assessment Act would mandate federal reviews of frontier models accessed by adversarial nations . [[5]](https://www.tomshardware.com/tech-industry/artificial-intelligence/trump-administration-reportedly-reviving-push-to-ban-chinese-ai-models-following-kimi-k3-launch-citing-cybersecurity-concerns-downloadable-open-weights-could-make-an-outright-u-s-ban-nearly-impossible-to-enforce-amid-growing-adoption)Treasury Secretary Bessent has taken the hardest line from within the administration. "If we see, especially, that overseas models are stealing from our great companies, we have the ability to sanction them because of this theft," he said, citing watermarks from American-built models found in Chinese AI systems [3]. The White House has also reportedly considered executive action that would require companies using Chinese AI to provide security guarantees and accept liability in the event of a breach . [[2]](https://fortune.com/2026/07/22/jensen-huang-chinese-open-source-ai-models-kimi-deepseek-washington-ban-nvidia-chips-data-centers-security/)Enforcement remains the central challenge. As multiple analysts and technologists have noted, open-weight models are freely downloadable from the internet, making an outright ban nearly impossible to implement in practice [7]. ## Why It Matters Huang's comments carry unusual weight because of Nvidia's central position in the AI supply chain. The company provides the GPUs that power training and inference for virtually every major AI lab — American and Chinese alike — and remains subject to US export controls that restrict the chips it can sell to China. His argument that open-source AI, including Chinese-origin models, expands the total market for compute infrastructure directly serves Nvidia's business interests, a tension he did not shy away from acknowledging [1]. The debate also exposes a rift between AI infrastructure companies, which benefit from broad adoption regardless of model provenance, and frontier AI labs like OpenAI and Anthropic, which have lobbied for tighter restrictions on models they allege were trained using distillation from their own outputs. Huang's counsel to Washington — "Let Anthropic run" — suggests he believes competition, not protection, is the right strategy [3]. Nvidia shares have risen roughly 29% year-to-date, giving the company a market capitalization north of $5 trillion, as investors continue to price in multi-year demand for AI compute infrastructure [6]. ## Companies mentioned ## Further sources [[1] Axios exclusive interview: Nvidia's Jensen Huang defends Chinese AI amid Kimi p… ↗](https://www.axios.com/2026/07/22/nvidia-jensen-huang-china-open-source-ai) [[2] Fortune: Jensen Huang on Chinese open-source AI models, Kimi, DeepSeek, and Was… ↗](https://fortune.com/2026/07/22/jensen-huang-chinese-open-source-ai-models-kimi-deepseek-washington-ban-nvidia-chips-data-centers-security/) [[3] Stocktwits: Nvidia CEO says US companies should 'absolutely' use Chinese AI mod… ↗](https://stocktwits.com/news-articles/markets/equity/nvda-jensen-huang-american-companies-absolutely-use-chinese-ai-models-bessent-sanctions-threat/cZZm0d1R7wU) [[4] Tom's Hardware: China's 2.8-trillion-parameter Kimi K3 beats Claude Fable 5 in … ↗](https://www.tomshardware.com/tech-industry/artificial-intelligence/moonshot-releases-2-8-trillion-parameter-kimi-k3) [[5] Tom's Hardware: Trump administration reportedly reviving push to ban Chinese AI… ↗](https://www.tomshardware.com/tech-industry/artificial-intelligence/trump-administration-reportedly-reviving-push-to-ban-chinese-ai-models-following-kimi-k3-launch-citing-cybersecurity-concerns-downloadable-open-weights-could-make-an-outright-u-s-ban-nearly-impossible-to-enforce-amid-growing-adoption) [[6] MarketBeat: NVIDIA (NASDAQ:NVDA) shares up 2.3% on July 22, 2026 ↗](https://www.marketbeat.com/instant-alerts/nvidia-nasdaqnvda-shares-up-23-time-to-buy-2026-07-22/)+1 more The stories that matter, in one email. Free — unsubscribe anytime.