OpenAI联合创始人Andrej Karpathy加入Anthropic，曾任特斯拉AI负责人
美国当地时间5月19日，OpenAI联合创始人之一、前特斯拉AI负责人安德烈·卡帕西（Andrej Karpathy）宣布加盟人工智能初创公司Anthropic。
在其个人社交平台账号上，安德烈·卡帕西写道，“更新一下个人近况：我已经加入了Anthropic公司。我认为未来几年在大语言模型（LLMs）前沿的工作将尤其具有塑造性和深远影响。我非常高兴能加入这个团队并重返研发工作。我对教育依然怀有深厚的热情，并计划在适当的时候恢复在这方面的工作。”
随后，Anthropic公司预训练团队负责人尼克拉斯·约瑟夫（Nicholas Joseph）转发了安德烈·卡帕西的推文并表示：“热烈欢迎安德烈·加入预训练团队。他将组建一支新团队，聚焦于利用Claude来加速预训练研究。我觉得没有任何人比他更适合干这个，非常期待我们合作。”
消息一出，美国人工智能领域多位业内人士在其推文下评论。英伟达具身智能研究负责人范麟熙（Jim Fan）评论道：这比当天举行的 Google I/O 开发者大会 Keynote 是“更重磅的新闻”；而 Anthropic公司麾下的Claude Code作者鲍里斯·切尔尼（Boris Cherny）等一众开发者圈里的核心人物，则迅速转发推文并表示热烈欢迎。
安德烈·卡帕西是AI领域里一位极具影响力的人物，不论是在自动驾驶还是在大模型的前沿技术开发过程中都曾发挥过关键作用，被媒体称为“卡神”。
公开资料显示，安德烈·卡帕西曾师从AI领域知名教授李飞飞，在斯坦福创立了传奇的 CS231n 课程。他是OpenAI最早的研究科学家，是公司最初的11名创始成员之一。2015年至2017年专注于深度学习与计算机视觉，2017年出任特斯拉AI总监，2022年从特斯拉离职，于2023年短暂回到OpenAI，约一年后再次离开，2024年创办AI教育公司Eureka Labs。
在特斯拉的五年中，安德烈·卡帕西力排众议，主导Autopilot视觉团队，构建了被称作“数据引擎”的闭环系统，把纯视觉（Vision-Only）自动驾驶神经网络架构推到数百万辆量产车上，成为将理论转化为工程现实的顶级架构师。
安德烈·卡帕西也是大模型时代的底层布道者。他早年提出的“软件 2.0（Software 2.0）”概念，即用神经网络权重取代传统手写代码，如今已成为行业共识。而在创办Eureka Labs期间，他引领了面向开发者的“AI 原生”教育。
Anthropic是一家成立于2021年的AI公司，公司创始人达里奥·阿莫迪（Dario Amodei）也曾任OpenAI研究副总裁。2026年Anthropic估值迅速攀升，接近1万亿美元，成为AI领域新晋巨头。其核心产品Claude系列大模型在企业级市场表现突出，企业客户占比超65%。据报道，Anthropic公司年化营收预计不久将突破450亿美元；而2025年其年化营收仅约90亿美元。
据悉，Anthropic公司计划2026年底启动IPO。5月15日，据财联社报道，Anthropic公司已就新一轮300亿美元的融资达成最终条款。本轮融资由龙舟投资（Dragoneer Investment Group）、绿橡资本（Greenoaks Capital）、红杉资本（Sequoia Capital）和 Altimeter Capital 联合领投，每家领投机构预计出资不低于20亿美元。多家现有投资方及新入局机构参与跟投，预计最快将于本月内完成交割。
本轮融资将使Anthropic的投后估值飙升至约9000亿美元。这不仅超过了OpenAI约8520亿美元的估值，也将使Anthropic成为全球估值最高的私营科技公司之一。
Anthropic公司从2023年开始便进入融资加速期，全球科技巨头与顶级VC的持续加码，推动其估值快速攀升。2023年初，谷歌率先向Anthropic投资3亿美元，获得约10%股权；同年5月，Spark Capital领投4.5亿美元C轮融资，估值升至500亿美元，谷歌、Salesforce Ventures等机构跟投；9月，亚马逊宣布初始投资12.5亿美元，承诺后续总投资规模达40亿美元。
2024年至2025年，亚马逊、谷歌持续追加投资，其中亚马逊累计投资达80亿美元，谷歌追加投资超10亿美元。
2025年3月，Lightspeed Venture Partners领投35亿美元E轮融资，Anthropic估值升至615亿美元。同年9月，Anthropic公司完成130亿美元F轮融资，估值达1830亿美元。2026年2月完成300亿美元G轮融资，估值升至3800亿美元。4月23日，其一级半市场股票交易隐含估值达近1万亿美元，首次反超OpenAI。
除融资外，Anthropic公司也在持续拓展算力基建合作，近期已与太空探索技术公司 SpaceX、谷歌、博通、亚马逊云科技等企业签约，锁定长期算力资源。据测算，未来数年Anthropic公司将为这类算力采购协议支付数千亿美元的成本支出。
当地时间5月18日，Anthropic公司宣布收购软件开发工具包（SDK）和MCP服务器工具领域企业Stainless，但未对外披露交易细节。据Anthropic公司介绍，Stainless成立于2022年，该公司将API规范转化为TypeScript、Python、Go、Java、Kotlin等多种语言的SDK；自Anthropic API早期阶段起，就为所有官方Anthropic SDK的生成提供技术支持。
本文来自界面新闻
OpenAI Co-founder Andrej Karpathy Joins Anthropic, Former Head of Tesla AI
On May 19 local U.S. time, Andrej Karpathy, co-founder of OpenAI and former head of Tesla AI, announced his joining of Anthropic, an artificial intelligence startup.
On his personal social media account, Karpathy wrote: "A quick personal update: I have joined Anthropic. I believe work at the cutting edge of large language models (LLMs) will be uniquely formative and influential in the years ahead. I’m thrilled to join the team and get back to research and engineering. I still hold a deep passion for education and plan to resume related work in due course."
Shortly afterwards, Nicholas Joseph, head of the Pre-training Team at Anthropic, reposted Karpathy’s post and commented: "A huge welcome to Andrej joining our pre-training team. He will lead a new group focused on advancing pre-training research leveraging Claude. I can think of no one better suited for this role, and I’m greatly looking forward to our collaboration."
The news soon sparked widespread discussions among U.S. AI industry insiders. Jim Fan, head of embodied intelligence research at NVIDIA, remarked that this was "bigger news" than the keynote speech delivered at Google I/O held the same day. Meanwhile, core figures in the developer community including Boris Cherny, creator of Claude Code at Anthropic, quickly shared the announcement and extended their warm welcome.
Andrej Karpathy is an iconic and highly influential figure in the AI sector, having played a pivotal role in both autonomous driving research and cutting-edge large language model development, earning him the nickname "God Karpathy" among industry insiders.
According to public profiles, Andrej Karpathy studied under renowned AI professor Fei-Fei Li and created the legendary CS231n course at Stanford University. He was one of the earliest research scientists at OpenAI and among its original eleven founding members. From 2015 to 2017, he dedicated himself to deep learning and computer vision research. In 2017, he took office as Tesla’s AI Director before stepping down from Tesla in 2022. He made a brief return to OpenAI in 2023 and departed again roughly one year later, then founded AI education firm Eureka Labs in 2024.
During his five-year tenure at Tesla, Karpathy overcame widespread doubts and led the Autopilot vision team. He built a closed-loop system known as the "data engine", and promoted the Vision-Only autonomous driving neural network architecture for mass application in millions of production vehicles, establishing himself as a top architect capable of translating theoretical technologies into practical engineering achievements.
Karpathy is also a pioneer who popularized core concepts in the large model era. He first put forward the notion of Software 2.0, which advocates replacing traditional manually written code with neural network weights, and this idea has now become an industry consensus. During his time running Eureka Labs, he spearheaded developer-oriented AI-native education initiatives.
Founded in 2021, Anthropic counts Dario Amodei, former Vice President of Research at OpenAI, among its founders. The company has seen its valuation soar rapidly in 2026, edging close to one trillion US dollars and emerging as a new heavyweight player in the AI landscape. Its flagship Claude series large models enjoy outstanding performance in the enterprise market, with enterprise clients accounting for over 65% of its user base. Reports predict that Anthropic’s annualized revenue will soon exceed 45 billion US dollars, a sharp surge from around 9 billion US dollars recorded in 2025.
It is reported that Anthropic plans to launch its IPO by the end of 2026. As quoted by Cls.cn on May 15, the company has finalized terms for a new round of 30 billion US dollar financing. This financing round is jointly led by Dragoneer Investment Group, Greenoaks Capital, Sequoia Capital and Altimeter Capital, with each lead investor committing no less than 2 billion US dollars. Multiple existing shareholders and new institutional investors will also participate in the follow-on investment, and the transaction is expected to be completed as early as within this month.
Upon completion of this round of financing, Anthropic’s post-money valuation will jump to approximately 900 billion US dollars, surpassing OpenAI’s valuation of around 852 billion US dollars and securing its place among the world’s most valuable private tech companies.
Anthropic has accelerated its financing pace since 2023, with continuous capital injections from global tech giants and top-tier venture capital firms fueling its rapid valuation growth. In early 2023, Google led a 300-million-US-dollar investment and secured roughly a 10% equity stake in the firm. In May the same year, Spark Capital spearheaded a 450-million-US-dollar Series C financing round, pushing its valuation to 50 billion US dollars, with Google and Salesforce Ventures joining as follow-on investors. In September, Amazon announced an initial 1.25-billion-US-dollar investment, pledging a total investment of up to 4 billion US dollars in the long run.
From 2024 to 2025, Amazon and Google kept increasing their investments in Anthropic. To date, Amazon’s total investment has reached 8 billion US dollars, while Google has injected over 1 billion US dollars in additional funds.
In March 2025, Lightspeed Venture Partners led a 3.5-billion-US-dollar Series E financing round, lifting Anthropic’s valuation to 615 billion US dollars. In September of the same year, it closed a 13-billion-US-dollar Series F financing deal, pushing its valuation up to 1.83 trillion US dollars. Another 30-billion-US-dollar Series G financing finished in February 2026 boosted its valuation to 3.8 trillion US dollars. On April 23, its implied valuation reflected by stock transactions in the secondary private market neared one trillion US dollars, overtaking OpenAI for the first time.
Apart from fundraising, Anthropic has also been actively expanding cooperation to consolidate computing power infrastructure. It has recently signed strategic agreements with SpaceX, Google, Broadcom, AWS and other major enterprises to secure long-term stable computing resources. Industry estimates suggest that Anthropic will spend hundreds of billions of US dollars in total on such computing power procurement agreements in the coming years.
On May 18 local time, Anthropic announced its acquisition of Stainless, a company specializing in software development kits (SDKs) and MCP server tools, without disclosing specific transaction details. Founded in 2022, Stainless is capable of converting API specifications into SDKs compatible with multiple programming languages including TypeScript, Python, Go, Java and Kotlin. It has provided core technical support for the development of all official Anthropic SDKs ever since the early stage of Anthropic’s API business.
Source: Jiemian News
返回首页