黃仁勳首發X挺開源AI：50企業聯署反封殺令 | Jensen Huang's X Debut Backs Open-Weight AI

7月 25, 2026

By Kit 小克 | AI Tool Observer | 2026-07-26

🇹🇼 黃仁勳首發X挺開源AI：50企業聯署反封殺令

黃仁勳在 2026 年 7 月 24 日發出他人生第一篇 X 貼文，內容不是財報、不是新晶片，而是一封聯署信：《Open Weights and American AI Leadership》（開源權重與美國 AI 領導地位）。這封信主張開源 AI 模型對美國的安全、創新與技術自主同樣重要，不該把國家 AI 戰略全部押在少數幾個封閉頂尖模型上。短短一天內，聯署企業從最初約 25 家暴增到 50 家，也讓「開源 vs 封閉」這場老爭論再度成為本週最熱話題。

這封信到底在吵什麼

背景是華盛頓正在評估是否限制或封殺中國開源模型（例如 DeepSeek、Qwen、Kimi 系列）在美國市場的使用。黃仁勳先前已多次公開表態，認為美國企業應該有自由選用中國開源模型的權利，他甚至說中國模型的競爭力反而會擴大市場對 Nvidia 晶片與資料中心基礎設施的需求，而不是威脅美國公司。這次的聯署信，等於是把這個立場正式化、集體化。

誰簽了、誰沒簽

初版聯署名單就頗有份量：

Nvidia、Microsoft、Meta、Palantir、IBM、Dell
Mistral、Hugging Face、Mozilla、Linux Foundation
Andreessen Horowitz、Y Combinator、CrowdStrike、Perplexity

隔天名單擴大到 50 家，補進了 OpenAI、Google、AMD、Cisco、Cloudflare、GitHub、Block、Ollama 等。但值得注意的是，Amazon 與 Anthropic 兩家至今都沒有簽署——尤其 Anthropic 一貫在 AI 安全議題上採取較保守立場，這次選擇不表態，某種程度上說明了業界對「開源是否等於安全」仍有分歧。

對開發者與一般用戶意味著什麼

先講白話：這封信不會立刻改變你手上任何一個 API 的定價或授權條款，它是政策遊說信，不是產品公告。但它反映幾個實際趨勢，值得追蹤：

開源模型的地位被主流巨頭正式背書，未來企業採購 AI 時「只能用封閉模型」的壓力可能會減少
若華盛頓真的祭出對中國開源模型的限制，會直接影響用 DeepSeek、Qwen 系列做本地部署或微調的開發者
Nvidia 的立場很現實——不管模型是開源還是封閉、美系還是中系，只要有人在跑推論，就要買它的晶片

對一般讀者來說，短期內不用做任何動作；但如果你的產品線壓在特定開源模型上，接下來一兩個月的政策動向值得多留意，別等到規則變了才臨時找替代方案。

好不好用，試了才知道。

🇺🇸 Jensen Huang's X Debut Backs Open-Weight AI

On July 24, 2026, Jensen Huang posted on X for the very first time — and instead of a chip announcement, it was a policy letter titled "Open Weights and American AI Leadership." The letter argues that open-weight AI models are just as critical to US safety, innovation, and technological sovereignty as closed frontier models, and that American AI strategy shouldn't rest on a handful of proprietary systems. Within 24 hours, the signatory list doubled from roughly 25 to 50 companies, making this the AI policy story everyone is talking about this week.

What the Letter Is Actually About

The backdrop: Washington is weighing restrictions on Chinese open-weight models like DeepSeek, Qwen, and Kimi. Huang has repeatedly argued that American companies should be free to use whatever models perform best, including Chinese ones — and that strong Chinese open models actually expand demand for Nvidia's chips and data center infrastructure rather than threaten US firms. This letter formalizes that position into a collective industry stance.

Who Signed — and Who Didn't

The original signatory list already carried real weight:

Nvidia, Microsoft, Meta, Palantir, IBM, Dell
Mistral, Hugging Face, Mozilla, Linux Foundation
Andreessen Horowitz, Y Combinator, CrowdStrike, Perplexity

By the next day the list had grown to 50, adding OpenAI, Google, AMD, Cisco, Cloudflare, GitHub, Block, and Ollama. Notably, Amazon and Anthropic remain absent from both versions of the list — Anthropic in particular has historically taken a more cautious stance on AI safety, and its silence here signals the industry still hasn't reached consensus on whether "open" automatically means "safer."

What This Actually Means for Developers

Bottom line: this letter changes zero API pricing or licensing terms today — it's a lobbying document, not a product launch. But it points to real trends worth tracking:

Open-weight models just got mainstream corporate backing, which may ease future pressure to default to closed-only procurement policies
If Washington does restrict Chinese open-weight models, it will directly affect teams self-hosting or fine-tuning DeepSeek or Qwen-family models
Nvidia's incentive is simple: open or closed, American or Chinese, every model that runs inference needs GPUs — and that's the business Huang actually cares about

If your stack doesn't depend on a specific open-weight model, there's nothing to do right now. But if it does, keep an eye on the policy timeline over the next month or two so you're not scrambling for a replacement after the rules change.

好不好用，試了才知道。

Sources / 資料來源

延伸閱讀 / Related Articles

AI 工具觀察站 — 每日精選 AI Agent 與工具趨勢
AI Tool Observer — Daily curated AI Agent & tool trends

搜尋此網誌

AI小貼士