内容来源：https://techcrunch.com/2025/09/30/former-openai-and-deepmind-researchers-raise-whopping-300m-seed-to-automate-science/

内容总结：

近日，由谷歌DeepMind前材料团队负责人埃金·多古斯·丘布克与OpenAI前研究副总裁利亚姆·费杜斯共同创立的Periodic Labs正式结束隐匿模式，宣布获得3亿美元种子轮融资，投资方包括安德森·霍洛维茨、DST全球、英伟达、Accel等顶级风投，以及杰夫·迪恩、埃里克·施密特和杰夫·贝索斯等科技界领袖。

该公司核心团队汇聚了人工智能与材料科学领域的顶尖研究者，曾主导开发OpenAI的ChatGPT、万亿参数神经网络、微软MatterGen材料发现模型等突破性项目。其愿景是打造"AI科学家"与自动化实验室，通过机器人自主进行物理实验、收集数据并持续优化，最终实现科学发现的自动化进程。

目前实验室首要目标是研发性能更优、能耗更低的新型超导体，并系统性收集实验过程中产生的物理世界数据，以突破当前大模型依赖网络数据训练的局限。这一方向也吸引了包括Tetsuwan Scientific、多伦多大学加速联盟在内的多家机构布局，标志着AI驱动科学发现的新趋势正在形成。

中文翻译：

周二，Periodic Labs结束隐匿模式正式亮相，并以3亿美元种子轮融资作为启动资金，其投资方堪称科技行业全明星阵容：安德森·霍洛维茨、DST、英伟达、Accel、埃拉德·吉尔、杰夫·迪恩、埃里克·施密特和杰夫·贝索斯。

Periodic Labs由埃金·多古斯·丘布克和利亚姆·费德斯共同创立。丘布克曾领导谷歌大脑和DeepMind的材料与化学团队，其主导项目包括名为GNoME的人工智能工具。研究人员表示，该工具在2023年发现了超过200万种新型晶体材料，这些材料未来或将为新一代技术提供动力。费德斯是OpenAI前研究副总裁，也是参与创建ChatGPT的研究人员之一，还曾带领团队创建了首个万亿参数神经网络。

这家初创公司的精干团队同样汇聚了参与过其他重大AI与材料科学项目的研究人员，涵盖从构建OpenAI智能体Operator到开发微软MatterGen（一个用于材料科学发现的LLM人工智能系统）等领域。

公司宣称，Periodic Labs的宏伟目标是实现科学发现自动化，创造AI科学家。这意味着要建立由机器人进行物理实验、收集数据、迭代优化并持续自主学习的自动化实验室。

该实验室的首要目标是发明新型超导体，期望其性能优于现有超导材料且可能更节能。但这家资金充裕的初创公司还希望发现其他新型材料。另一项目标是收集AI科学家在探索新材料过程中，通过混合、加热及其他方式处理各种粉末与原材料所产生的全部物理世界数据。

【Disrupt 2025科技大会邀请】汇聚万名科技与风投领袖，共促成长与人脉拓展。Netflix、Box、a16z、ElevenLabs、Wayve、Hugging Face、埃拉德·吉尔、维诺德·科斯拉——届时将有250多位行业领袖主导200多场专题会议，为您提供驱动初创企业成长与提升竞争优势的深度洞察。切勿错过TechCrunch二十周年盛会，向科技领域顶尖声音学习。立即预购门票最高立省444美元。

公司在介绍性博文中指出：“迄今为止，科学AI的进步主要来自基于互联网数据训练的模型”，而大语言模型已“耗尽”了可获取的网络资源。“在Periodic，我们正在构建AI科学家及供其运作的自主实验室。”他们期待这些实验室不仅能发明下一代材料，更能产出AI模型进化所需的全新宝贵数据。

尽管这可能是为实现该目标组建的最杰出研究团队之一，但并非唯一从事AI科学家研发的机构。自2023年起，AI作为化学发现自动化工具已成为学术研究课题，包括Tetsuwan Scientific等小型初创公司，以及Future House非营利组织和多伦多大学加速联盟等机构也致力于此领域。

英文来源：

Periodic Labs came out of stealth on Tuesday with a war chest of $300 million as a seed round, backed by a tech industry who’s who: Andreessen Horowitz, DST, Nvidia, Accel, Elad Gil, Jeff Dean, Eric Schmidt, and Jeff Bezos.
Periodic Labs was founded by Ekin Dogus Cubuk and Liam Fedus. Cubuk led the materials and chemistry team at Google Brain and DeepMind, where one of his projects was, for instance, an AI tool called GNoME. That tool discovered over 2 million new crystals in 2023, materials that could one day be used to power new generations of technology, researchers say.
Fedus is a former VP of Research at OpenAI, and one of the researchers who helped create ChatGPT. He also led the team that created the first trillion-parameter neural network.
Its small team is likewise filled with researchers who have worked on other major AI and materials science projects, from building OpenAI’s agent Operator to working on Microsoft’s MatterGen, an LLM materials science discovery AI.
The goal of Periodic Labs is nothing less than to automate scientific discovery, creating AI scientists, the company says. This means building labs where robots conduct physical experiments, collect data, iterate, and try again, learning and improving as they go.
The lab’s first goal is to invent new superconductors that it hopes perform better and possibly require less energy than existing superconducting materials. But the well-funded startup also hopes to find other new materials.
Another goal is to collect all the physical world data that its AI scientists produce as they mix and heat and otherwise manipulate various powers and raw materials in their search for something new.
Join 10k+ tech and VC leaders for growth and connections at Disrupt 2025
Netflix, Box, a16z, ElevenLabs, Wayve, Hugging Face, Elad Gil, Vinod Khosla — just some of the 250+ heavy hitters leading 200+ sessions designed to deliver the insights that fuel startup growth and sharpen your edge. Don’t miss the 20th anniversary of TechCrunch, and a chance to learn from the top voices in tech. Grab your ticket before doors open to save up to $444.
Join 10k+ tech and VC leaders for growth and connections at Disrupt 2025
Netflix, Box, a16z, ElevenLabs, Wayve, Hugging Face, Elad Gil, Vinod Khosla — just some of the 250+ heavy hitters leading 200+ sessions designed to deliver the insights that fuel startup growth and sharpen your edge. Don’t miss a chance to learn from the top voices in tech. Grab your ticket before doors open to save up to $444.
“Until now, scientific AI advances have come from models trained on the internet” and LLMs have “exhausted” the internet as a source that can be consumed, the company says in an introductory blog post. “[A]t Periodic, we are building AI scientists and the autonomous laboratories for them to operate.”
The hope is that, not only will the labs invent next-generation materials, but they will produce invaluable fresh data that AI models can consume to continue their evolution.
While this might be one of the most impressive groups of researchers to assemble a startup for this purpose, it’s not the only one working on AI scientists. AI as a tool to automate chemistry discoveries has been a topic of academic research since at least 2023. It is the pursuit of tiny startups like Tetsuwan Scientific, as well as nonprofits like Future House and the University of Toronto’s Acceleration Consortium.