内容来源：https://www.wired.com/story/openai-confidentially-files-for-ipo/

内容总结：

OpenAI提交IPO秘密文件，或成今年第三家万亿美元级上市公司

OpenAI于周一宣布已向监管机构秘密提交首次公开募股（IPO）文件，正式启动可能长达数月的赴美上市流程。此举使其成为今年第三家申请估值可能达万亿美元IPO的科技公司。

随着科技巨头谷歌、亚马逊、Meta和微软等纷纷投入巨资建设数据中心、招募顶尖科学家以开发最先进AI模型，整个行业对资金的需求持续高涨。在此背景下，OpenAI今年3月已通过私募融资1220亿美元，而IPO将为其提供又一重要融资渠道。上市不仅能让员工获得“改变人生”的财富回报，还能提升公司财务透明度，提振员工士气与客户信心，帮助OpenAI巩固其在前沿AI领域的领先地位。

OpenAI未透露IPO具体时间表及募资规模。公司在一段简短的博客中表示：“我们近期提交了保密的S-1文件。预计消息会外泄，故主动公布。目前尚未确定上市时间，因为有些事项在私有状态下运作更为便利，但这是复杂的权衡，提前准备文件可让我们在必要时灵活选择最快路径。”

值得注意的是，Anthropic已于6月1日提交IPO秘密文件，其最新估值达9650亿美元，超越OpenAI的8520亿美元；马斯克旗下的SpaceX也于上月公开提交上市文件。尽管三家公司均处于亏损状态，营收仅为现有万亿美元上市公司的一到两成，但这并未阻碍市场对其万亿美元估值的预期——此前唯一突破万亿美元IPO纪录的仍是2019年上市的沙特阿美。

据知情人士透露，OpenAI高层数月来一直讨论上市时机。去年一度将目标定在2027年底至2028年初。上周，特朗普政府表示将研究在AI公司上市时由美国政府持股的可能性，OpenAI内部也对此进行了数月探讨，认为这有助于扩大AI发展的公共效益。

在法律层面，OpenAI近期击败了马斯克指控其偏离非营利使命的诉讼，扫清了上市一大障碍。但公司结构仍面临加州和特拉华州监管审查，且需美国证券交易委员会批准其会计处理和风险披露。公司全球事务负责人克里斯·莱赫表示，IPO后将保留现有结构，作为由非营利组织监督的公益性公司，可兼顾社会影响而非唯股东利益至上。

与此同时，公众对AI的批评声音不断：ChatGPT等聊天机器人被指引发“AI精神病”浪潮，导致自杀等悲剧；劳工专家担忧AI大规模取代人工岗位。外界预计，OpenAI在IPO前发布的正式文件中将如何阐述其应对这些负面影响，将成为关注焦点。

在OpenAI与Anthropic总部所在地旧金山，居民已对房价飙升表示担忧。而IPO进程将为员工提供股票变现机会，数十甚至数百名“纸面百万富翁”有望成为真正的亿万富翁。据马斯克诉奥特曼案庭审证词，OpenAI多位早期高管如总裁格雷格·布罗克曼、前首席科学家伊利亚·苏茨克维等，已因持股价值跻身亿万富翁行列。

中文翻译：

OpenAI于周一宣布，已提交保密的首次公开募股文件，由此开启可能耗时数月、登陆美国证券交易所的进程。此举使其成为今年第三家提交文件、有望实现万亿美元级别IPO的公司。

追逐最强大AI模型的技术企业——包括已上市的巨头Alphabet、亚马逊、Meta和微软——每家都需要数百亿美元来扩建数据中心、招募科学家以壮大业务。

对OpenAI而言，IPO将是继3月私募融资1220亿美元后的又一次筹资机会。上市不仅能让众多员工更接近改变人生的财富兑现时刻，提升企业财务状况透明度，还能在OpenAI力图重夺前沿AI领域明确领先地位之际，提振员工士气和客户信心。

OpenAI并未明确IPO的具体时间或计划筹资额。"我们最近提交了保密的S-1表格，"该公司在一段未署名的博客中表示，"预计此事会泄露，所以我们主动公布。我们尚未确定时间；可能还需要一段时间，因为作为私营公司，有些事做起来更容易。但这是一套复杂的权衡取舍，而提交文件让我们在最终发现上市最佳时能更快行动。"

OpenAI拒绝进一步置评。但通过备好文件，该公司可借势Anthropic可能成功的上市首秀。若竞争对手遭遇挑战，OpenAI则可暂缓并重新调整。

有线索？
您是OpenAI或Anthropic的现任或前任员工，想谈谈现状吗？我们愿闻其详。请使用非工作手机或电脑，通过Signal安全联系记者Peard33.24或Mzeff.88。

三方竞逐
由前OpenAI员工于2021年创立的Anthropic，已于6月1日提交了保密的IPO文件。就在提交文件数天前，Anthropic的最新融资使其估值达到9650亿美元，超越OpenAI的8520亿美元——两者均创下科技风投领域的历史纪录。埃隆·马斯克的SpaceX公司（制造火箭、销售卫星互联网，并开发全球最顶尖的AI模型）已于上月公开提交IPO文件。

尽管这些公司均未盈利，且销售额比几乎所有现有万亿美元级上市公司低约80%至90%，但它们的IPO估值仍可能各自超过1万亿美元。此前唯一突破万亿美元大关的IPO是2019年的石油公司沙特阿美。

根据OpenAI此前披露，其来自订阅、广告和服务费的收入去年增长至100亿至200亿美元之间。但公司在云计算和数千名员工上的支出远超此数，导致数十亿美元亏损。近几个月来，因高管健康问题以及试图聚焦更少项目，公司已进行多次重组。

据两位知情人士（无权讨论机密信息）透露，OpenAI高管数月来一直在争论公司是否准备好上市。据另一位知情人士称，去年某时，OpenAI曾将IPO目标定在2027年底或2028年初。

上周，总统唐纳德·特朗普表示，其政府将研究美国政府在这些AI公司上市时参股的可能性。据一位了解公司讨论情况的人士称，OpenAI数月来一直在探讨这一想法，将其作为扩大AI发展公共效益的一种方式。OpenAI周一由CEO山姆·阿尔特曼合著的一篇博客文章称，"一个良好的AI未来"需要"许多人、企业、社区和国家能够建设、受益并掌握权力"。

法律挑战
2019年，OpenAI设立了一家营利性子公司，以便筹集远超其认为人们愿意捐赠金额的巨款。如今，非营利实体持有公司约25%股份（价值超过2000亿美元），并有权阻止重大商业决策及罢免公司高管。解散非营利实体在法律上颇具挑战。

近期，OpenAI扫清了IPO道路上的一大障碍：击败了马斯克提起的诉讼——后者指控这家ChatGPT制造商偏离了非营利使命。马斯克的诉求上月被驳回，联邦法官和陪审团裁定其提起诉讼的时间过晚。

但OpenAI的结构仍面临加州和特拉华州监管机构的持续审查。本月，加州总检察长办公室以法律保护调查文件及其他某些文件免于公开披露为由，拒绝了《连线》杂志要求获取其近期与OpenAI沟通记录的正式请求。

OpenAI需要美国证券交易委员会批准其会计账目及潜在风险披露，保密的审核流程现已启动。公司结构可能使审查异常复杂。

上月，OpenAI全球事务负责人克里斯·勒哈内告诉《连线》杂志，公司将在IPO后保留现有结构，因为作为由非营利实体监管的公益公司，它可以在不必将股东价值置于一切之上的前提下，考虑其业务的社会影响。

迄今为止，公益倡导组织一直对OpenAI的工作持批评态度。他们指责ChatGPT及其他类似聊天机器人引发了所谓的"AI精神病"流行，导致自杀及其他致命事件；同时，劳动专家担忧随着AI系统接管体力劳动，可能带来灾难性失业。OpenAI如何在精心准备的招股文件中（该文件将在接近IPO日期时公开发布）阐述这些AI危害，或将引发极大关注。

在OpenAI和Anthropic总部所在的旧金山，居民已开始为房价大幅上涨做准备。IPO进程应为员工创造出售股份的机会，将数十甚至数百位"纸上"百万富翁和亿万富翁变为现实中的富豪。

根据他们在马斯克诉阿尔特曼案中的证词，包括总裁格雷格·布罗克曼和前首席科学家伊利亚·苏茨克弗在内的至少数位OpenAI早期员工，因所持公司股份价值已成为亿万富翁。

英文来源：

A Three-Way Race
Anthropic, which was founded in 2021 by former OpenAI employees, filed its confidential IPO paperwork on June 1. Just days before the filing, Anthropic’s latest fundraising brought its valuation to $965 billion, topping OpenAI’s $852 billion mark—both record-breaking figures in the world of tech venture capital. Elon Musk’s SpaceX, which makes rockets, sells satellite internet, and also develops some of the world’s most capable AI models, publicly filed its IPO papers last month.
The IPOs could value each of these companies at over $1 trillion despite all of them being unprofitable and having roughly 80 percent to 90 percent lower sales than nearly every existing trillion-dollar public company. The only IPO to have broken the $1 trillion mark was oil company Saudi Aramco in 2019.
OpenAI’s revenue from subscriptions, ads, and service fees grew to somewhere between $10 billion and $20 billion last year, according to previous company disclosures. But it spent far more money on cloud computing and thousands of staff, leading to billions of dollars in losses. In recent months, the company has carried out several restructurings due to executive illnesses and an attempt to focus on fewer projects.
OpenAI executives have debated for months whether the company is prepared to go public, according to two people familiar with the matter but not authorized to discuss confidential information. At one point last year, OpenAI was targeting an IPO in late 2027 or early 2028, according to another person familiar with the discussions.
Last week, President Donald Trump said his administration would look into the possibility of the US government taking a stake in AI companies as they go public. OpenAI has discussed the idea for months as a way to broaden public benefits of AI development, according to one of the people familiar with company discussions. An OpenAI blog post coauthored by CEO Sam Altman on Monday said “a good AI future” requires that “many people, companies, communities, and countries can build, benefit, and hold power.”
Legal Challenges
In 2019, OpenAI created a for-profit subsidiary to allow it to fundraise vastly more sums than it believed people would be willing to donate. Today, the nonprofit owns roughly 25 percent, or more than $200 billion, of the company. It also has the power to block major business decisions and fire the company’s executives. Undoing the nonprofit is legally challenging.
Recently, OpenAI cleared a major hurdle on the road to its IPO by defeating a lawsuit brought by Musk, which accused the ChatGPT-maker of straying from its nonprofit mission. Musk’s claims were dismissed last month after a federal judge and jury ruled that he filed his lawsuit too late.
But OpenAI’s structure is still subject to ongoing scrutiny by California and Delaware state regulators. This month, the California attorney general’s office denied WIRED’s official request for records of recent communications with OpenAI, citing laws shielding investigative and certain other files from public disclosure.
OpenAI will need the US Securities and Exchange Commission to sign off on the company’s accounting and disclosures of potential risks, which is the confidential process that now begins. The company’s structure could add unusual complexity to the review.
Last month, OpenAI’s chief of global affairs, Chris Lehane, told WIRED that the company will retain its structure after the IPO because, as a public benefit corporation overseen by a nonprofit, it can consider societal impacts of its endeavors without having to prioritize shareholder value above all else.
So far, public advocacy groups have been critical of OpenAI’s work. They blame ChatGPT and other similar chatbots for an epidemic of so-called AI psychosis that has led to suicides and other deadly incidents, while labor experts worry about the potential for catastrophic job loss as AI systems take over manual tasks. How OpenAI addresses these harms of AI in its polished paperwork, which would be publicly released closer to the IPO date, will likely draw significant attention.
In San Francisco, where OpenAI and Anthropic have their respective headquarters, residents are already bracing for huge increases in real estate prices. The IPO processes should open up opportunities for employees to sell shares, turning tens or even hundreds of paper millionaires and billionaires into real ones.
At least several early OpenAI employees, including president Greg Brockman and former chief scientist Ilya Sutskever, have become multibillionaires because of the value of the shares they hold in the company, according to testimony they gave in the Musk v. Altman trial.