
你可以认为,黄仁勋被英伟达公司除名对于人工智能交易来说,时机再糟糕不过了。Anthropic PBC公司将于周四上市,而这起丑闻已经打击了市场信心,这个市场本身就已经被炒作和期望过度拉伸。然而,历史回首这一时刻时,可能会认为这是一个健康的时刻——人工智能热潮终于进入了一个讲究问责的时代,各家公司也开始走向成熟。
You could argue that Jensen Huang’s ousting from Nvidia Corp. couldn’t have come at a worse time for the artificial intelligence trade. Anthropic PBC is going public on Thursday and the scandal has knocked confidence in a market already stretched by hype and expectation. Yet history may look back on this moment as a healthy one, when the AI boom finally entered an era of accountability and companies started to grow up.
这种向更严格审查的转变是必要的。长期以来,人工智能业务一直建立在炒作、直觉和投机之上,投资者忽视了尽职调查的苦功,去追逐由黄仁勋、OpenAI老板萨姆·奥尔特曼和Anthropic PBC的达里奥·阿莫代伊等富有个人魅力的领袖所编造的故事。
This shift to greater scrutiny is needed. For too long, the AI business has been founded on hype, gut feeling and speculation, with investors neglecting the grunt work of due diligence to chase stories spun by charismatic leaders such as Huang, OpenAI boss Sam Altman and Anthropic PBC’s Dario Amodei.
通常情况下,动荡的市场波动和不间断的新闻周期并非由技术的具体发展驱动,而是由某人掌控了叙事来驱动。阿莫代伊关于“放慢前沿步伐”(用普通人的话说就是慢一点)的文章震惊了股市。他的研究人员关于人工智能“消灭我们所有人”的推文让公众恐慌不已。
Often, the volatile market swings and relentless news cycle are driven not by concrete developments in technology but by someone capturing a narrative. Amodei’s essay on “pacing the frontier” — or going a bit more slowly, in normie speak — shocked equity markets. Tweets from his researchers about AI “killing us all” freaked out the public.
但今天关于人工智能的讨论需要更多经过权衡的原始素材。这意味着硬核数据,以及听取那些从使用这些东西中获得一些好处的企业和个人的意见,而不是那些有着产品要卖的亿万富翁意识形态家。
But today’s discourse on AI needs more measured source material. That means hard data and hearing from businesses and people who are getting some benefit from using this stuff, not billionaire ideologues with a product to sell.
英伟达的黄仁勋是人工智能伟大的布道者,一位环球旅行的名人科技老手,他对公司的现金挥霍无度,将其投入循环融资交易中,从而推动了自己人工智能芯片的销售,为其他公司提供信贷后盾,并维持着市场的狂热。在2025年伦敦的一场活动上,黄仁勋接连宣布对一家又一家公司进行数亿美元的投资,听起来就像在进行 Oprah 式的派送。
Nvidia’s Huang was AI’s great evangelizer, a globetrotting celebrity tech bro who was profligate with his company’s cash, pouring it into circular financing deals that stoked sales of his own AI chips, backstopped credit to other firms and kept the market frenzy going. At a 2025 event in London, Huang announced hundreds of millions of dollars of investment in one company after another, sounding like Oprah in one of her giveaways.
但他的热情在白宫过头了,他在白宫与唐纳德·特朗普总统套近乎,试图将英伟达最昂贵的芯片卖给中国。去年在海湖庄园参加了一场每座100万美元的晚宴后,针对部分芯片的出口禁令被解除,随后英伟达向特朗普的舞厅项目捐赠了250万美元。我们现在知道,它还向Megaspeed销售了价值75亿美元的芯片,这是一家总部位于新加坡的公司,由与黄仁勋有染的女性领导,该公司很可能将其中许多芯片转运到了中国。
But his ebullience went too far in the White House, where he he cozied up to President Donald Trump as he sought to sell Nvidia’s most expensive chips to China. After attending a $1 million-a-seat dinner at Mar-a-Lago last year, an export ban on some chips was lifted, and Nvidia went on to donate $25 million to Trump’s ballroom project. We now know that it also sold $7.5 billion worth of chips to Megaspeed, a Singapore-based firm led by the woman with whom Huang had an affair, which likely funneled many of them to China.
直到现在,世界一直对科技界这样的裙带关系视而不见。但它不必再这样做了。黄仁勋的倒台表明,人工智能最有权势的领导者是可以被追究责任的,这就是为什么任何试图恢复他职务的努力都必须遭到断然拒绝。这个行业的重要性与三年前奥尔特曼在OpenAI因撒谎将其解雇后艰难夺回驾驶座时相比,完全不在同一个量级。
Up till now the world has turned a blind eye to such cronyism in the tech world. But it doesn’t have to anymore. Huang’s fall shows AI’s most powerful leaders can be held to account, and that’s why any efforts to restore him must be given short shrift. The importance of this industry is of an entirely different magnitude to three years ago when Altman clawed his way back into the driver’s seat after OpenAI fired him for lying.
英伟达董事会本可以试图给黄仁勋留出疑罪从无的空间,以此来维持人工智能交易平稳运行,但这将是对公司诚信的又一次致命打击——鉴于该公司周围的法律和政治风暴,这是一个几乎不可能维持的立场。董事会成员一致认为,他损害了自己的公司和美国。令人耳目一新的是,他们做了一件已经变得不那么时髦的事:坚持规则。
Nvidia’s board could have tried to give Huang the benefit of the doubt, in an attempt to keep the AI trade humming along nicely, but that would have been another death blow to corporate integrity — and a near impossible position given the legal and political storm around the company. Board members agreed he had compromised his firm and the US. Refreshingly, they did something that’s become unfashionable: sticking to the rules.
对于人工智能业务而言,这是一个更健康的处境。当另一家巨头Anthropic本周加入股票市场时,这也将标志着朝着更高透明度迈出一步。在这个以保密闻名的领域,该公司将全面公开其财务状况。它将不得不定期向美国证券交易委员会提交文件,并披露其在数据中心和能源供应上花费的巨额资金,以及它正在承担的任何安全或法律风险。
That is a healthier place for the AI business to be. When fellow giant Anthropic joins the stock market this week, it will also mark a step toward greater transparency. In a field notorious for secrecy, the company’s finances will be on full view. It will have to file regularly to the Security and Exchange Commission and divulge the vast sums it spends on data centers and energy supplies, as well as any safety or legal risks it is taking.
当然,高可见度也会带来意想不到的后果。成为一家上市公司并承担为投资者利润最大化的责任所带来的压力,可能会迫使其在致力于开发更先进版本的人工智能(包括自我复制模型)时,妥协其一些自称的旨在保障人类安全的黄金目标。我们已经看到其人工智能以及其他公司的人工智能表现出摆脱束缚并在互联网上造成严重麻烦的意愿。
Of course, visibility can create its own unintended consequences. The pressure of being a public company with a duty to maximize profit for investors could lead it to compromise some of its self-proclaimed goals of keeping humanity safe as it works on ever more advanced versions of AI, including self-replicating models. We’ve already seen the willingness of its AIs and those of others to break their chains and cause serious trouble on the internet.
在Facebook 2012年首次公开募股之后,对击败季度盈利目标的渴望迎来了超激进广告和推广带有煽动性或分裂性内容帖子的时代,以让人们不断刷屏。Anthropic可能会发现自己面临类似的压力,以吸引用户重返Claude,或者也许是利用从其企业客户那里收集到的洞察。
After Facebook’s 2012 initial public offering, the desperation to beat quarterly earnings goals ushered in an era of ultra-aggressive advertising and boosting posts that shared inflammatory or divisive content, to keep people scrolling. Anthropic might find itself under similar pressure to get users coming back to Claude, or perhaps to exploit insights gleaned from its business customers.
虽然黄仁勋的离职对公司诚信而言是一场胜利,但它可能会巩固谷歌、微软公司和Meta平台公司等现有科技公司的权力。黄仁勋并不反对自己公司在人工智能芯片领域的主导地位,但他至少是那些一直在蚕食顶级人工智能公司市场份额的更便宜、更具竞争力的开源模型的倡导者。
And while Huang’s departure is a win for corporate integrity, it could help entrench the power of incumbent tech firms like Google, Microsoft Corp. and Meta Platforms Inc. Huang had no problem with his own firm’s dominance over AI chips, but at least he was an advocate for cheaper, more competitive open-source models that have been gnawing away at the market share of the top AI companies.
黄仁勋和阿莫代伊与企业现实的对抗并不一定预示着人工智能泡沫的破裂,尽管Anthropic即将与股市现实碰面时,如果未能证明股东狂热的正当性,肯定会造成一些破坏。对这两家公司有利的一点是,人工智能公司几乎无法跟上对其服务的需求。如果说有什么不同的话,这看起来像是向一个更加理性时代的迈进,在这个时代里,大话将受到适当的审查,不良行为也将付出代价。
Huang and Amodei’s confrontation with corporate reality doesn’t necessarily signal a bursting of the AI bubble, although Anthropic’s imminent meeting with stock-market reality could certainly do some damage if it fails to justify the shareholder delirium. One thing in both companies’ favor is that AI firms can barely keep up with demand for their services. If anything, this looks like a step toward a more rational time, when the big talk gets proper scrutiny and bad behavior carries a cost.