IPO 在即,Anthropic 发布新 AI 模型与办公工具

这家生成式 AI 厂商发布了面向企业工作的新功能,并对模型进行了大幅更新,为其历史性的公开上市做准备。
Esther Shittu,AI Business 新闻撰稿人
Anthropic 位于旧金山 Howard 街的总部。Benjamin Fanjoy / Stringer via Getty Images
尽管股市出现经济动荡,Anthropic 仍计划以 1.33 万亿美元的估值提交 IPO,远低于此前预期的 2 万亿美元。
该计划让企业得以了解这家厂商的财务状况,尤其是在企业将这家生成式 AI 厂商最新更新的 Claude Fable 5.2 和 Claude Mythos 5.2 模型,以及 Claude Enterprise 功能整合进自身工作流之际。
Anthropic 在周二向 SEC 提交的修订版注册声明中表示,将以每股 70 至 78 美元的价格出售约 8.25 亿股股票。借助此次申报,Anthropic 预计将募资约 610 亿美元,并从芯片制造商 AMD 等投资者处再获得 260 亿美元。
这家厂商在其即将进行的公开发行中确定了股价,时值 Anthropic 发布 Claude Fable 5.2 和 Claude Mythos 5.2 的一天之后。
据 Anthropic 介绍,这些模型着重于编码、推理和视觉方面的改进,并且比过去的模型能更清晰地沟通。该厂商周二还发布了 Claude Enterprise 的新功能,涵盖会议、白板、电子表格、数据,以及面向 Adobe、Atlassian 和 HubSpot 的连接器。
此次发布以及该厂商持续推进 IPO 的计划,出现在金融市场经历一周动荡之后;在这轮动荡中,许多投资者对政治格局的变化以及人们对 AI 数据中心的普遍反感作出反应,纷纷抛售手中的科技股。
许多对冲基金在 11 月 3 日中期选举后立即抛售科技股,转而买入医疗保健和公用事业等板块的股票。
IPO 并不等于价值
然而,据 Futurum Equities 首席市场策略师 Shay Boloor 称,Anthropic 目前为其 IPO 预计的 1.3 万亿美元估值——IPO 预计于下周、即美国感恩节假期之前进行——不应被视为等同于该厂商的技术实力。
Boloor 表示:“投资者是在为这样的预期买单:Anthropic 能够维持超常增长,同时将更多营收转化为利润。”
他补充说,1.3 万亿美元的估值表明,前沿 AI 的价值终于开始与股市的实际表现挂钩,而不是与投资者对市场走向的预期挂钩。
不过 Boloor 表示,Anthropic 持续推进公开上市对企业而言是好事,因为这让人们得以一窥该公司的财务状况。
他继续说道:“这带来了更高的财务透明度和资本获取渠道,当企业将 Claude 嵌入到他们预期要依赖多年的工作流程中时,这一点尤为重要。”
但他认为,IPO 并不会改变 AI 市场的竞争格局,因为企业仍将基于成本、性能和可靠性,从 Anthropic 以及其美国 AI 模型竞争对手 Google 和 OpenAI 中选择模型。
Anthropic 已证明自己能够持续留住企业客户。该公司披露,截至 9 月 30 日,其净收入留存率为 900%。Boloor 提醒说,这一数字虽然亮眼,但不应将其解读为那些早已把该技术嵌入自身工作流的企业客户的留存率。
他说,真正令人瞩目的是,这家 Claude 的开发商通过一项按 IPO 价格进行的单独私募配售,从 Google、Nvidia、AMD 以及其他 AI 算力供应商那里获得了 260 亿美元。
Boloor 表示,这“是极大的信心背书,但也凸显出 AI 生态在财务上已变得多么相互交织——因为这些公司一边投资 Anthropic,一边又从 Anthropic 需要采购的庞大算力基础设施中获益”。
这种持续的未来支出也从台积电(TSMC)的最新消息中得到印证——台积电为全球市场代工大部分 GPU,并计划分别向 OpenAI 和 Anthropic 各投资 5 亿美元。
企业工作环境
与此同时,Anthropic 持续向投资者证明,它不仅仅是一家模型提供商,其最新为 Claude Enterprise 增添的功能尤其说明了这一点。
Claude Meetings 让 Claude 能够加入视频通话,并在团队仍在讨论时就着手工作,由此进军历来属于 Microsoft 和 Google 的企业工作空间领域。
Claude Whiteboards 和 Claude Sheets 让团队能够借助共享白板和电子表格,使用 Claude 智能体进行构建。Claude Data 则把公司的数据转化为仪表板、数据故事和视频。
“它让 Claude 更贴近业务意图产生的起点,”Tekonyx 创始人兼分析师 Sid Nag 表示。“Anthropic 实际上是在争夺企业工作界面。”
Nag 表示,随着微软和谷歌的应用中涌现出许多新能力,Anthropic 更大的机会在于让 Claude 从决策的起点就将其捕捉下来。
Nag 表示,对 Anthropic 而言,在其公开上市前的几天里推出这类能力,表明它不只是一家纯粹的 AI 开发者。
“如今,该公司正从一家只是生产 LLM(大语言模型)以及提示词、引擎和智能体的 AI 实验室,转向真正把这一切整合起来,打造一套生产力工作流,”他说。
但 Anthropic 通过发布 Fable 5.2 和 Mythos 5.2 仍然表明,其最有效的产品依然是模型。
“它们更快、也更便宜,”Informa TechTarget 旗下 Omdia 的分析师 Lian Jye Su 表示。
Claude Mythos 5.2 和 Claude Fable 5.2 的价格为每百万输入 Token 8 美元、每百万输出 Token 40 美元。这一新定价相较这两款模型的 5.1 版本有明显下降——后者为每百万输入 Token 10 美元、每百万输出 Token 50 美元。
Anthropic 表示,Fable 5.2 能用比 Fable 5.1 更少的 Token 完成具有挑战性的任务。据该厂商介绍,Claude Mythos 5.2 是其面向网络安全工程能力最强的模型,发现的漏洞数量多于此前各代模型。
“如今我们看到网络安全威胁大幅增加,Anthropic 再次强调这一领域很有帮助,”Su 表示。“就 AI 解决方案能带来的其他实际收益而言,这是业界少数能达成共识的事情之一。”
新闻撰稿人,AI Business
Esther Shittu 自 2021 年起持续报道 AI 技术与行业趋势。作为 Targeting AI 播客的联合主持人,她与专家、思想领袖和从业者对话,探讨 AI 领域的关键进展。在加入 AI Business 之前,她曾为 SearchEnterpriseAI、New York Daily News、Bklyner 和 Brooklyn Daily Eagle 撰稿。不深入钻研 AI 世界时,她会把时间投入到自己热爱的项目中,并照顾三个女儿。
译文由英文原文机器翻译,术语按行业惯例统一,可在文末展开原文核对。
查看英文原文(With IPO looming, Anthropic launches new AI models, workplace tools)
The generative vendor released new capabilities targeting enterprise work and significantly updated models as it prepares for the historic public offering.
Esther Shittu,News Writer, AI Business
Anthropic's headquarters on Howard Street in San Francisco. Benjamin Fanjoy / Stringer via Getty Images
Despite the economic upheaval in the stock market, Anthropic plans to file its IPO at a valuation of $1.33 trillion, well below the previously expected $2 trillion.
The plan gives enterprises insight into the vendor's financials, especially as they integrate the generative AI vendor’s newly updated Claude Fable 5.2 and Claude Mythos 5.2 models, as well as Claude Enterprise capabilities, into their workflows.
Anthropic said in an amended registration statement with the SEC filing on Tuesday that it would sell about 825 million shares for $70 to $78 per share. With the filing, Anthropic expects to raise about $61 billion and an another $26 billion from investors such as chipmaker AMD.
The vendor set the price of its upcoming public shares a day after Anthropic released Claude Fable 5.2 and Claude Mythos 5.2.
The models focus on improvements in coding, reasoning and vision and can communicate more clearly than past models, according to Anthropic. The vendor also released new capabilities on Tuesday for Claude Enterprise for meetings, whiteboards, spreadsheets, data and connectors to Adobe, Atlassian and HubSpot.
The releases and the vendor’s continued plans to pursue an IPO come after a week of upheaval in the financial markets, during which many investors responded to the shift in the political landscape and widespread dislike of AI data centers by selling their tech shares.
Many hedge funds sold tech stocks right after the Nov. 3mid-term elections, switching to stocks in sectors such as healthcare and utilities.
IPO does not equal worth
However, the $1.3 trillion that Anthropic now projects for its IPO - expected next week, before the U.S. Thanksgiving holiday - should not be seen as equivalent to the vendor’s technology, according to Shay Boloor, chief market strategist at Futurum Equities.
“Investors are paying for the expectation that Anthropic can sustain extraordinary growth while converting more of that revenue into profits,” Boloor said.
He added that the $1.3 trillion valuation shows that the value of frontier AI is finally being priced against how the stock market is faring, rather than what investors think the market will do.
Nevertheless, Anthropic’s continued progress toward a public offering is good for enterprises because it provides a glimpse into the company’s finances, Boloor said.
“It brings greater financial transparency and access to capital, which matters when companies are embedding Claude into workflows they expect to rely on for years,” he continued.
But the IPO does not change the competitive dynamics in the AI market because enterprises will continue to choose models from Anthropic, as well as its U.S. AI model rivals Google and OpenAI, based on cost, performance and reliability, he argued.
Anthropic has demonstrated that it continues to retain its enterprise customers. It disclosed a 900% net revenue retention rate as of Sept.30. While impressive, Boloor cautioned against interpreting it as a retention rate for enterprises that have long embedded the technology in their workflows.
What is impressive is the $26 billion that the Claude maker is getting from Google, Nvidia, AMD, and other providers of AI compute with a separate private placement at the IPO price, he said.
Boloor said this is a “tremendous vote of confidence but also highlights how financially interconnected the AI ecosystem has become since these are companies investing in Anthropic while simultaneously benefiting from an enormous amount of computing infrastructure Anthropic needs to purchase.”
The continued future spending is also evident in the latest news from Taiwan Semiconductor Manufacturing Company, which fabricates most GPUs for the world market and plans to invest $500 million each in OpenAI and Anthropic.
The enterprise work environment
Meanwhile, Anthropic continues to prove to investors that it is more than just a model provider, especially wth its latest additions to Claude Enterprise.
Claude Meetings lets Claude join a video call and start work in what has traditionally been Microsoft's and Google's enterprise workspace territory, while the team is still talking.
Claude Whiteboards and Claude Sheets enable teams to build with the Claude agent using shared boards and spreadsheets. Claude Data turns a company's data into dashboards, data stories and videos.
"It puts Claude closer to where the business intent starts," said Sid Nag, founder and analyst at Tekonyx. "Anthropic is really competing for the enterprise work interface."
With many new capabilities available across Microsoft and Google applications, Anthropic's larger opportunity lies in enabling Claude to capture decisions from their beginning, Nag said.
For Anthropic, launching capabilities such as these in the days leading up to its public offering shows that it is not only a pure AI developer, Nag said.
"Now the company is moving away from just being another AI lab that produces LLMs (large language models) and prompts, engines and agents to actually taking all those things together and creating a productivity workflow," he said.
But Anthropic still showed that its most effective products are models, with the release of Fable 5.2 and Mythos 5.2.
“They are faster, cheaper,” said Lian Jye Su, an analyst at Omdia, a division of Informa TechTarget.
Claude Mythos 5.2 and Claude Fable 5.2 cost $8 per million input tokens and $40 per million output tokens. This new pricing is a noticeable reduction from version 5.1 of the two models, which cost $10 per million input tokens and $50 per million output tokens.
Anthropic said Fable 5.2 can accomplish challenging tasks with fewer tokens than Fable 5.1. Claude Mythos 5.2 is the vendor’s most capable model for cybersecurity engineering and catches more vulnerabilities than previous models, according to the vendor.
“We are seeing a massive increase in cybersecurity threats nowadays, and having Anthropic reemphasize this domain is helpful,” Su said. “It is one of the few things that the industry can agree on in terms of other actual benefits that AI solutions can bring.”
News Writer, AI Business
Esther Shittu has covered AI technologies and industry trends since 2021. As co-host of the Targeting AI podcast, she talks with experts, thought leaders and practitioners exploring critical AI developments. Before AI Business, she wrote for SearchEnterpriseAI, the New York Daily News, Bklyner and the Brooklyn Daily Eagle. When she's not diving deep into the world of AI, she spends her time on passion projects and raising her three daughters.
原文为英文,正文由自动抓取并保留原文。
原文发布于 AI Business(With IPO looming, Anthropic launches new AI models, workplace tools),标题、摘要与正文由机器翻译,术语按行业惯例统一,未经人工逐句校订。