行业多源确认84°

Anthropic IPO 首日大涨 34.3%,完全稀释估值约 2 万亿美元

精选理由

Anthropic 上市当天股价涨了 34.3%,估值冲到约 2 万亿。这篇文章有意思的地方是讲清楚了 OpenAI 抢在上市前发模型为什么没用——市场看的是叙事和估值锚,不是 Astra 的科学基准成绩。关心两家融资格局的人应该看看。

Anthropic 于今日完成 IPO,开盘价 111.50 美元,较 83 美元发行价高出 34.3%,完全稀释估值约 2 万亿美元,IPO 加同步私募预计带来 1048 亿美元净募资。这些资金的主要用途之一是继续租赁数据中心,使 SpaceX 旗下算力业务成为间接受益方。文中认为真正的输家是 OpenAI:无论 GPT-6.2 Astra 和 Sol 的实际表现如何,Anthropic 已经率先建立公开估值基准,OpenAI 未来若 IPO 将被拿来对标。

正文

作者还分析了 OpenAI 在 Anthropic IPO 前两天发布新模型的策略:若 OpenAI 在上市前表现足够强势,本可以分流投资者资金。但市场对这次发布的观感不佳,多数人仍认为 Opus 5.5 和 Fable 5.2 是最好的模型,Astra 在科学等领域领先也未能扭转公众认知,发布反而强化了 Anthropic 的上市表现。

原文 · kimmonismus

A quick take on Anthropic’s IPO and how it ties in with OpenAI.

Anthropic made history today, and I think it’s safe to say the IPO was a success across the board. Its fully diluted valuation reached about $2 trillion at the start of trading. Shares opened at $111.50, 34.3% above the $83 IPO price, and stayed above the IPO price in subsequent trading.

Anthropic got everything it could have hoped for from its IPO and expects $104.8 billion in net proceeds from the IPO and the concurrent private placement. That’s fresh capital it needs and plans to use, in particular, to keep leasing data centers. That makes SpaceXSI an indirect beneficiary. In other words, Anthropic’s IPO is a win for both Anthropic and SpaceX.

OpenAI is the loser here. And even if a lot of people don’t want to hear it, I think it needs to be said. OpenAI lost in this case regardless (!) of whether GPT-6.2 Astra and Sol really are that bad or have gotten significantly better since then. The reason is that Anthropic’s IPO achieved what OpenAI is aiming for. With its IPO, Anthropic has gained a head start in terms of financing and established a public valuation benchmark against which OpenAI will be measured should it pursue an IPO of its own. That’s a fact.

The stronger OpenAI had looked ahead of the IPO and the more convincingly it had shown investors that it was the real play, the less money would have gone into Anthropic and the more capital would be available to OpenAI next year. That’s what matters.

The two are each other’s biggest competitors, and I can’t imagine a scenario in which OpenAI didn’t try, at least in some way, to undercut Anthropic’s valuation by putting out a release two days before Anthropic’s IPO. I think it’s pretty clear that it didn’t work. And (again) it makes no difference whether OpenAI’s model is good or bad.

What was disastrous about the release was the public perception. The impression was that it wasn’t a good release, and the market bore that out. As the IPO numbers show, it posed no threat to Anthropic.

If anything, it went the other way. In my view, OpenAI’s latest release probably hurt OpenAI and strengthened Anthropic, because for a lot of people, Opus 5.5 and Fable 5.2 are still the best models even after the release. And yes, I know Astra leads in many areas, including science. That carries little weight with the market or in the public eye. That’s how this IPO should be put in context. And that’s how the numbers should be read.

Big day for Anthropic, a lot of work for OpenAI. But the end is still open.

产业坐标与信源核验

本条情报归属 AITOP「行业」核心栏目,重点聚焦北美前沿实验室与硅谷科技巨头。在跨区域产业观察坐标中,该动态直接关联「中美大模型技术代差与商业化对照」研究主线,反映出当前技术演进与商业落地的最新风向。涉及核心实体与技术议题包括:Anthropic、OpenAI、IPO、行业融资,读者可通过下方标签进入对应专题页查阅历史脉络与关联对比。

信源核验说明:本文原始线索采编自「kimmonismus」(@kimmonismus),经 AITOP 多维评分管线结构化提炼。截至目前,全网已有 11 家独立信源对该事件进行交叉印证与跟进报道(含 小互、Rowan Cheung、IT之家 等),属于近期高关注度行业信号。