广告
加载中

张一鸣 又创造了一个100亿

李馨婷 2026-10-12 08:53
李馨婷 2026/10/12 08:53

邦小白快读

EN
全文速览

总1:关键新闻

1. 张一鸣旗下字节跳动分拆出的AI制药公司新生实验室完成首轮融资2.9亿美元(约20亿元人民币),投后估值约15亿美元(约100亿元人民币),字节保留56%股份,仍是控股方。

2. 这是今年国内AI制药领域公开披露金额最大的一笔融资,投资方包括红杉中国、IDG资本、高瓴创投、五源资本、高榕创投、春华创投以及中国生物制药等。

总2:实操提醒

1. AI制药确实有真实进展:AI介入后,新药研发从早期药物发现到确定临床前候选药物,周期从传统4-5年压缩到8-12个月。

2. 但别轻信“AI治愈癌症”口号。目前多数AI制药项目还停在临床前和早期临床,三期项目极少;真正衡量标准是能否通过临床试验并治好患者。

3. 全球资本正在涌入,近10年AI制药领域投资超400亿美元;普通人要关注的是这些钱能不能转化为安全有效的新药,而不是只看融资数字。

总1:品牌进入高壁垒领域的方式

1. 字节跳动不是简单以主品牌做药,而是把内部AI4S团队拆分成新生实验室,独立融资并保留控股权;同时用红杉中国、IDG、高瓴、中国生物制药等豪华股东做品牌背书。

2. 这种做法能降低主品牌直接试错的风险,又借助外部资本和产业资源提升新品牌可信度,是科技品牌延伸至医疗健康的参考案例。

总2:产品研发与竞争定价

1. 新生实验室已有真实产品管线:发布生物分子结构预测模型Protenix、SeedFold,蛋白质结合剂设计模型PXDesign、SeedProteo,并披露首个全谱IL-17小分子抑制剂,瞄准银屑病、强直性脊柱炎等自身免疫疾病。

2. 行业估值很贵:Xaira成立时种子轮超10亿美元,Isomorphic Labs首轮外部融资6亿美元、B轮达21亿美元。AI制药的品牌价值由技术、资本和产业合作共同定价。

3. 风险提醒:现在AI制药的三期项目很少,品牌宣传一旦过度承诺疗效,会损伤长期信任;产品能否通过临床验证才是品牌护城河。

总1:增长市场与机会

1. 国内AI制药今年融资金额已超157亿元,超过去3年总和;头部机构频繁出手,金沙江联合资本出手6次、启明创投5次,说明这条赛道正处于资本密集进入期。

2. 创业者可借鉴字节的孵化模式:先在平台内部完成技术验证,再分拆为独立公司引入外部融资并保留控制权,降低早期风险、放大增长可能。

总2:参与方式与风险提示

1. 文章显示专业人才正从大厂外溢:顾全全、肖文之等人离开字节,在AI制药、蛋白设计、生物基础模型和药物发现平台方向创业,并获头部美元机构多轮加注;有相关能力的团队可关注这一窗口。

2. 合作方式上,医药产业资本(如中国生物制药)已出现在新生实验室投资人名单中,说明产业方在寻求与AI制药公司绑定,卖家可寻找配套服务、数据、临床资源等卡位机会。

3. 风险在于落地成果不足:AI制药项目多数处于临床前和早期临床,III期极少;不要只看融资热度去押注单一药企或单一管线,应优先切入解决研发效率的刚需环节。

总1:制造与生产需求

1. 新生实验室正在推进IL-17小分子抑制剂等临床前项目,药物一旦进入临床和上市,小分子药物的工艺放大、质量控制和规模化生产是后续刚需,制药工厂可提前关注这类新锐公司的量产合作机会。

2. 文章强调AI蛋白质设计模型能按治疗需求设计新蛋白质,意味着候选药物种类和生产要求会更多样,工厂需要储备柔性生产、连续制造等能力,以匹配更快的研发节奏。

总2:数字化与合作启示

1. 字节AI4S团队用AI重构科学计算和药物发现流程,说明工厂数字化不是简单买系统,而是围绕研发周期、工艺优化等真实瓶颈建设数据模型能力。

2. 中国生物制药等产业资本参与新生实验室融资,提示工厂可以主动与AI制药公司形成资本或业务绑定,切入从研发到生产的产业化节点。

3. 同时要重视风险:AI制药目前离大规模商业化还有距离,工厂在产能投入上应分期规划、与临床进度挂钩,避免盲目扩产。

总1:行业趋势与客户痛点

1. 趋势:AI制药赛道被资本追捧,国内今年融资超157亿元,超过前3年总和;AI介入后新药早期研发周期从4-5年压缩到8-12个月。药企和AI制药公司都有提高效率、缩短周期的压力。

2. 痛点:Nature Reviews Drug Discovery文章指出,AI应用多集中在临床前阶段,III期项目寥寥;AI早期决策和最终临床影响脱节,是行业根本挑战之一。服务商可围绕临床转化、试验设计、真实世界数据提供支持。

总2:新技术与服务机会

1. 技术高地集中在生物分子结构预测(Protenix、SeedFold)和蛋白质设计(PXDesign、SeedProteo)等模型,服务商可提供算力、数据服务、评测体系、模型部署等配套。

2. 科技巨头分拆趋势带来客户群变化:字节将AI制药团队独立融资并保留控股,顾全全等人才也出来创业;服务商要关注从大厂分拆的创业公司,它们更大概率愿意采购外部专业服务。

3. 资本结构上,红杉中国、IDG、高瓴、中国生物制药等都在入场,意味着服务商面对的客户资金实力更强,同时对交付质量和临床价值验证的要求也会更高。

总1:平台孵化与治理模式

1. 字节跳动孵化AI制药的路径:内部Seed团队投入技术研发,积累模型能力后再分拆成新生实验室,引入红杉中国、IDG、高瓴创投、五源资本、高榕创投、春华创投和中国生物制药等资本,同时保留56%控股。

2. 对平台商的启示:前沿领域投入大、周期长,平台可以用分拆融资的方式分散风险,同时利用控股权保持战略协同,避免创新业务拖累主干。

总2:运营管理与风险规避

1. 平台需要敢于让内部人才独立发展:顾全全、肖文之等核心成员离开字节创业后反而获得头部美元机构加注,平台与其限制人才,不如通过投资或合作把外溢变成生态连接。

2. 风险规避:新生实验室的估值虽高,但AI制药真正的临床验证还很少,三期项目寥寥;平台在布局类似新赛道时,不能只看融资热度,应为技术失败和长期研发预留空间。

3. 平台可以借鉴其“技术验证—产业合作—独立融资”节奏,而不是过早大额自担所有商业化成本。

总1:产业新动向与商业模式

1. 科技巨头进入AI制药出现新组织模式:字节把内部AI4S团队分拆为新生实验室并独立融资,自己保留56%股份;同时核心科学家顾全全、肖文之等离开字节创业,形成“大厂孵化—人才外溢—创业获投”的产业链闭环。

2. 资本动向:国内AI制药融资超157亿元,超过此前3年总和;全球近10年投资超400亿美元,头部项目如Xaira、Isomorphic Labs估值高涨,研究可关注这一轮资本配置的效率与泡沫风险。

总2:技术问题与政策启示

1. 技术进展不可忽视:SeedFold宣称全面超越AlphaFold3,SeedProteo性能超过AlphaProteo和RFdiffusion;但这些评测多由团队自己公布,需要独立验证和标准化基准研究。

2. 核心问题:AI应用主要停留在临床前阶段,III期项目极少;I期成功率90%的归因也可能因项目建立在成熟生物学基础上而高估。研究应聚焦AI决策与临床结果之间的脱节。

3. 政策建议:需要建立AI制药临床疗效的证据标准、临床前数据披露要求和从早期发现到III期转化的激励机制,避免资本口号掩盖真实疗效。

返回默认

声明:快读内容全程由AI生成,请注意甄别信息。如您发现问题,请发送邮件至 run@ebrun.com 。

我是 品牌商 卖家 工厂 服务商 平台商 研究者 帮我再读一遍。

Quick Summary

Summary 1: Key News

1. Newborn Laboratory, an AI drug discovery company spun out of ByteDance (founded by Zhang Yiming), completed its first financing round of $290 million (approximately RMB 2 billion), at a post-investment valuation of about $1.5 billion (approximately RMB 10 billion). ByteDance retains 56% ownership and remains the controlling shareholder.

2. This is the largest publicly disclosed financing in China's AI drug discovery sector this year. Investors include Sequoia Capital China, IDG Capital, Hillhouse Ventures, 5Y Capital, Gaorong Ventures, Chunhua Ventures, and China Biopharmaceutical Corporation.

Summary 2: Practical Reminders

1. AI drug discovery has made real progress: with AI intervention, the timeline from early drug discovery to identifying a preclinical candidate has been compressed from the traditional 4-5 years to 8-12 months.

2. But don't take the "AI cures cancer" slogan at face value. Currently, most AI drug discovery projects remain in preclinical and early clinical stages, with very few Phase III trials. The real measure is whether drugs can pass clinical trials and cure patients.

3. Global capital is pouring in, with over $40 billion invested in AI drug discovery over the past decade. The public should focus on whether this money translates into safe and effective new drugs, not just the financing numbers.

Summary 1: How Brands Enter High-Barrier Fields

1. ByteDance did not simply build a drug business under its main brand. Instead, it spun out its internal AI4S team into Newborn Laboratory, raised independent financing, and retained controlling ownership. At the same time, it leveraged prestigious shareholders such as Sequoia Capital China, IDG, Hillhouse, and China Biopharmaceutical Corporation for brand endorsement.

2. This approach reduces the risk of direct trial-and-error under the main brand while using external capital and industry resources to enhance the credibility of the new brand. It is a reference case for tech brands extending into healthcare.

Summary 2: Product Development and Competitive Pricing

1. Newborn Laboratory already has a real product pipeline: it released biomolecular structure prediction models Protenix and SeedFold, protein binder design models PXDesign and SeedProteo, and disclosed its first pan-spectrum IL-17 small molecule inhibitor targeting autoimmune diseases such as psoriasis and ankylosing spondylitis.

2. Industry valuations are expensive: Xaira raised over $1 billion in seed funding at inception, Isomorphic Labs raised $600 million in its first external round and $2.1 billion in a Series B. The brand value of AI drug companies is priced by a combination of technology, capital, and industry partnerships.

3. Risk reminder: Phase III AI drug projects are still rare today. If brand promotion overpromises efficacy, it will damage long-term trust. Passing clinical validation is the real moat for the brand.

Summary 1: Growth Market and Opportunities

1. Domestic AI drug discovery financing has exceeded RMB 15.7 billion this year, more than the total of the previous three years combined. Leading institutions have been active, with JSD Capital making 6 deals and Qiming Venture Partners 5, indicating a period of intense capital influx into this track.

2. Entrepreneurs can learn from ByteDance's incubation model: first validate the technology internally within a platform, then spin it off into an independent company, bring in external financing while retaining control, reducing early-stage risk and amplifying growth potential.

Summary 2: Participation Paths and Risk Warnings

1. The article shows that specialized talent is spilling out of big tech: Gu Quanquan, Xiao Wenzhi, and others left ByteDance to start ventures in AI drug discovery, protein design, biological foundation models, and drug discovery platforms, attracting multiple rounds of investment from leading dollar funds. Teams with relevant capabilities should watch this window.

2. In terms of collaboration, pharmaceutical industry capital (e.g., China Biopharmaceutical Corporation) has already appeared on Newborn Laboratory's investor list, indicating that industry players are seeking to bind with AI drug companies. Sellers can look for positioning opportunities in supporting services, data, clinical resources, and more.

3. The risk lies in insufficient tangible outcomes: most AI drug projects are in preclinical and early clinical stages, with very few Phase III trials. Do not bet on a single drug company or pipeline solely based on financing heat. Instead, prioritize entering essential pain points that solve R&D efficiency.

Summary 1: Manufacturing and Production Demand

1. Newborn Laboratory is advancing preclinical projects such as IL-17 small molecule inhibitors. Once drugs enter clinical trials and reach market, process scale-up, quality control, and large-scale production for small molecule drugs become essential follow-up needs. Pharmaceutical factories should watch for future production collaboration opportunities with such emerging companies.

2. The article emphasizes that AI protein design models can create new proteins based on therapeutic needs, meaning that candidate drug types and production requirements will become more diverse. Factories need to build capabilities in flexible manufacturing and continuous processing to match faster R&D cycles.

Summary 2: Digitalization and Partnership Insights

1. ByteDance's AI4S team used AI to reconstruct scientific computing and drug discovery workflows, showing that factory digitalization is not simply buying systems, but building data modeling capabilities around real bottlenecks like R&D cycles and process optimization.

2. The participation of China Biopharmaceutical Corporation and other industrial capital in Newborn Laboratory's financing suggests that factories can proactively form capital or business ties with AI drug companies, entering the industrialization nodes from R&D to production.

3. At the same time, risks should be taken seriously: AI drug discovery is still far from large-scale commercialization. Factories should plan capacity investment in stages, linked to clinical progress, to avoid blind expansion.

Summary 1: Industry Trends and Client Pain Points

1. Trend: AI drug discovery is being pursued by capital. Domestic financing this year exceeds RMB 15.7 billion, more than the total of the previous three years. With AI, the early R&D cycle for new drugs has been compressed from 4-5 years to 8-12 months. Both pharma companies and AI drug firms face pressure to improve efficiency and shorten timelines.

2. Pain points: An article in Nature Reviews Drug Discovery notes that AI applications are mostly concentrated in the preclinical stage, with very few Phase III projects. The disconnect between AI-driven early decisions and ultimate clinical impact is one of the industry's fundamental challenges. Service providers can offer support in clinical translation, trial design, and real-world data.

Summary 2: New Technologies and Service Opportunities

1. Technology high ground is concentrated in biomolecular structure prediction (Protenix, SeedFold) and protein design (PXDesign, SeedProteo) models. Service providers can offer supporting capabilities such as computing power, data services, evaluation systems, and model deployment.

2. The trend of tech giants spinning off units brings changes in the client base: ByteDance spun out its AI drug team with independent financing and retained control, while talents like Gu Quanquan have also left to start ventures. Service providers should watch startups spun out of big tech, as they are more likely to purchase external professional services.

3. In terms of capital structure, Sequoia Capital China, IDG, Hillhouse, and China Biopharmaceutical Corporation are all entering the field. This means service providers face clients with stronger financial strength, but also higher demands for delivery quality and clinical value validation.

Summary 1: Platform Incubation and Governance Model

1. ByteDance's path to incubating AI drug discovery: its internal Seed team invested in technology R&D, accumulated model capabilities, then spun off into Newborn Laboratory. It brought in capital from Sequoia Capital China, IDG, Hillhouse Ventures, 5Y Capital, Gaorong Ventures, Chunhua Ventures, and China Biopharmaceutical Corporation, while retaining 56% control.

2. Implication for platform companies: in frontier fields with high investment and long cycles, platforms can use spin-off financing to diversify risk while maintaining strategic alignment through controlling ownership, avoiding the innovation business dragging down the core.

Summary 2: Operations Management and Risk Mitigation

1. Platforms need to allow internal talent to develop independently: core members like Gu Quanquan and Xiao Wenzhi left ByteDance to start ventures and instead attracted additional investment from leading dollar funds. Rather than restricting talent, platforms can turn outflow into ecosystem connections through investment or cooperation.

2. Risk mitigation: although Newborn Laboratory's valuation is high, true clinical validation in AI drug discovery remains scarce, with very few Phase III projects. When platforms deploy into similar new tracks, they should not rely solely on financing heat; they need to reserve room for technical failure and long-term R&D.

3. Platforms can learn from its rhythm of "technology validation - industry collaboration - independent financing" rather than prematurely bearing all commercialization costs at scale.

Summary 1: Industry Dynamics and Business Models

1. A new organizational model has emerged for tech giants entering AI drug discovery: ByteDance spun off its internal AI4S team into Newborn Laboratory with independent financing, retaining 56% ownership. Meanwhile, core scientists Gu Quanquan, Xiao Wenzhi, and others left ByteDance to start their own ventures, forming a closed-loop industry chain of "incubation by big tech - talent spillover - startups receiving funding."

2. Capital trends: Domestic AI drug financing exceeded RMB 15.7 billion, more than the previous three years combined. Globally, investment has exceeded $40 billion over the past decade. Flagship projects such as Xaira and Isomorphic Labs have seen soaring valuations. Research can examine the efficiency and bubble risk of this round of capital allocation.

Summary 2: Technical Questions and Policy Implications

1. Technological progress cannot be ignored: SeedFold claims to surpass AlphaFold3 across the board, and SeedProteo claims performance exceeding AlphaProteo and RFdiffusion. However, these evaluations are mostly self-reported by the team and require independent validation and standardized benchmarking research.

2. Core issue: AI applications remain mainly in the preclinical stage, with very few Phase III projects. A 90% Phase I success rate attribution may also be overestimated because projects are built on mature biology. Research should focus on the disconnect between AI decisions and clinical outcomes.

3. Policy recommendations: There is a need to establish evidence standards for clinical efficacy of AI drugs, preclinical data disclosure requirements, and incentive mechanisms for transitioning from early discovery to Phase III, to prevent capital slogans from masking real efficacy.

Disclaimer: The "Quick Summary" content is entirely generated by AI. Please exercise discretion when interpreting the information. For issues or corrections, please email run@ebrun.com .

I am a Brand Seller Factory Service Provider Marketplace Seller Researcher Read it again.

刚刚成为亚洲新晋首富的张一鸣,“一不小心”又创造了一个“最贵”。

近日,AI制药公司Anew Labs(新生实验室)完成首轮独立融资,融资总额2.9亿美元(折合人民币20亿),投后估值达15亿美元(折合人民币约100亿元)。按照目前公开披露的融资金额,这是今年国内AI制药领域金额最大的一笔融资。

新生实验室是今年6月从字节跳动分拆出的AI制药业务。本轮融资完成后,字节跳动保留56%股份,保持控股地位。

这是一轮阵容豪华的融资,涵盖了红杉中国、IDG资本、高瓴创投、五源资本、高榕创投、春华创投等头部投资机构,中国生物制药等多家医药产业资本也参与其中。

对于一家刚独立运营3个月的公司来说,20亿元的首轮融资高得梦幻。但放眼当下热火烹油的AI制药赛道,这样的融资并不算少。

据PitchBook报道,近10年间,风投机构们已向AI制药领域投资超过400亿美元。具体到国内,据投中嘉川CVsource统计,国内的164家AI制药企业中,共有117家获得投资。截至9月,今年AI制药行业的融资金额已经超过157亿元,超过此前3年的总和。这其中,金沙江联合资本与启明创投,今年的出手次数分别达到6次与5次。

再看从DeepMind拆分出来后、去年首轮外部融资达到6亿美元的AI制药公司Isomorphic Labs。

同样顶着“科技巨头孵化+AI制药”设定的新生实验室,拿到这笔融资似乎也算不上意外。

治愈疾病,是AI最诱人的应用场景之一,而AI制药能持续吸引资本涌入,某种程度上也是这份人类愿景分量的体现。只是,从目前的落地成果来看,这份愿景离真正兑现,还有很长一段路要走。

意料之内的拆分

从豆包、飞书,到剪映、小云雀,被戏称为AI工厂的字节跳动,无论杀进哪个互联网场景,都不会让人感到惊讶。但制药看起来是个例外,这离字节的主业太远了。

实际上,字节跳动在AI制药领域,已经早有布局。

新生实验室的前身,是字节跳动Seed旗下的AI for Science(AI4S)药物发现团队。Seed成立于2023年,字节跳动内部对它的定位,是“寻找通用智能的新方法,追求智能上限”。至于AI4S团队,则专注于科学计算领域,围绕生物领域基础模型、量子化学、分子动力学等方向,试图用AI推动科学领域的研究范式突破。

过去两年,字节AI4S团队在AI制药的模型能力上,已经攒下了不少成果。

2025年前后,字节AI4S团队陆续发布生物分子结构预测模型Protenix、SeedFold,以及蛋白质结合剂设计模型PXDesign和SeedProteo。简单来说,Protenix和SeedFold解决的是“看清生物分子的结构”,PXDesign和SeedProteo则试图按照治疗需求“设计出新的蛋白质”。

据字节AI4S团队前核心成员顾全全的描述,SeedFold是全球首个在广泛的基准测试和能力上全面超越AlphaFold 3的生物分子结构预测模型——AlphaFold 3是DeepMind在2024年发布的同类模型。此外,按照团队公布的测试结果,团队推出的另一蛋白质结合剂设计模型SeedProteo,性能同样超越了多个业内同类产品,包括DeepMind的AlphaProteo、2024年诺贝尔化学奖得主戴维·贝克(David Baker)团队开源的RFdiffusion。

产业化探索方面,今年4月,字节跳动在美国过敏、哮喘和免疫学会年会上,公开披露了一款处于临床前阶段的IL-17小分子抑制剂。

IL-17是一种关键的促炎细胞因子,在银屑病、强直性脊柱炎等疾病中驱动过度炎症反应。

字节这次披露的是全球首个“全谱”IL17小分子抑制剂,有望转化为银屑病和其他自身免疫性疾病的治疗方法。

字节的AI4S团队几乎已经摸到了AI制药产业端的门槛。但这也带出了另一个问题:如果字节要亲自下场做药,接下来要付出的探索成本,实在太高了。

医药研发遵循“双十定律”:耗时十年、投入十亿美元,验证周期以年为单位。这中间还要闯过实验体系搭建、管线推进、临床前风险、商业化等关卡,每一环都是和字节核心业务截然不同的行业逻辑。

或许是意识到了这个问题,AI4S团队的人才也在持续外溢。

据媒体6月报道,AI4S团队的顾全全与肖文之等多位成员已经离开或准备离开字节创业,方向集中在AI制药、蛋白设计、生物基础模型和药物发现平台,相关创业项目已经获得头部美元机构多轮加注。

这其中,顾全全是机器学习理论、大模型对齐以及AI4S科学智能领域的知名学者。2023年加入字节Seed前,他已经是加州大学洛杉矶分校计算机科学系的终身副教授。在字节AI4S团队,顾全全主要牵头AI药物研发工作,SeedFold与SeedProteo都出自他带领的团队。

到了9月10日,顾全全在X上宣布,他已经成立了一家全新的AI4S企业Geodesic Intelligence。公司的目标,是构建用于药物研发的通用人工智能,并找到从生物学到药物的最短路径。

在开始强调聚焦主干业务时,选择将AI制药团队拆分独立运营并保留控股权,对于字节跳动来说,显得顺理成章。

“AI治愈癌症”,还有多远?

纵观市场动态,字节跳动拆分团队入局AI制药并不突然。过去几年,依托AI4S成果进行新药研发,早已是全球科技巨头与国际药企的共同选择。

GPT时刻后,AI迅速颠覆了文本写作、视频生成、代码编写等数字业态。此前因为缺乏突破性技术而一度沉寂的AI制药,也再度迎来爆发点。

传统的新药研发,需要经历“靶点发现—先导化合物发现—分子优化—临床前研究—临床试验(共三期)”等环节。按老规矩,一款新药从研发到上市,平均耗时十年以上、10亿美元、成功率不到10%。一组发布于2023年、被广泛引用的行业数据显示,AI介入后,新药研发环节中,从早期药物发现到确定临床前候选药物的周期,已经能从传统的4到5年,压缩至8到12个月。

在“拯救生命”这个人类崇高命题下,无论是社会价值还是商业价值,AI制药赛道都具备庞大的想象力。资本纷纷下场押注也变得十分合理。

由诺贝尔化学奖得主戴维·贝克创立的端对端AI制药公司Xaira Therapeutics,2024年成立时就完成了超10亿美元种子轮融资。Isomorphic Labs在2025年完成6亿美元的首轮外部融资后,今年5月的B轮融资高达21亿美元。

乐观的情绪也蔓延到了AI行业顶层的话语体系。2025年时,Open AI创始人山姆·奥特曼(Sam Altman)在博客文章中大胆假设:“如果拥有10吉瓦的算力,AI或许就能找到治愈癌症的方法。”到了今年9月,奥特曼对AI的期待更进一步。在接受采访时他表示,“治愈癌症对AI来说还不够有野心,行业应该瞄准更大的目标。”

资本在狂欢,旗手在呐喊。那么,美丽的AI制药乌托邦,眼下成色几何?

今年8月,《Nature Reviews Drug Discovery》(自然综述:药物发现)发表了一篇关于AI药物研发现状与未来路径的综述文章。这篇由16位来自剑桥大学、哈佛医学院、阿斯利康、欧洲生物信息研究所等高校、药企与AI制药公司的作者联合撰写的文章,泼了几盆冷水。文章核心观点如下:

1、在新药研发过程中,临床疗效才是最关键的评定维度。但目前AI在小分子药物研发领域的应用主要集中在临床前阶段。在一众AI制药公司中,目前,仅有数十个项目处于I期和II期,而III期项目则寥寥无几。

2、近期一些基于少量数据点的文献显示,AI制药I期临床试验项目的成功率明显提高至90%,并将其归因于AI的潜在影响,但这些项目大多基于已建立的疾病生物学和化学基础,从而降低了在后续阶段(或更晚阶段)遇到安全问题的可能性。

3、药物研发过程中,可以基于代理数据使用AI进行的早期决策阶段,与药物通常在几年后才显现的临床影响之间的脱节,是药物发现以及量化人工智能对这一过程的影响所面临的根本挑战之一。

显然,尽管AI已经大幅压缩了新药研发周期,但在变数极大的生物化学领域,AI还是有些绕不过去的行业门槛。获得资本慷慨输血的AI制药企业们,还需要在很多环节向市场证明自己。

这么看来,在一众给AI制药站台的大佬里,比起奥特曼,他的对手——Anthropic创始人达里奥·阿莫迪(Dario Amodei),似乎更有自知之明。

2024年10月,阿莫迪发布了一篇博文。在谈到AI对生物医学的潜在影响时表示,阿莫迪表示,AI有望帮助开发针对癌症基因组特征的个性化治疗方案,未来癌症死亡率和发病率或许能够大幅下降,甚至有望消除大多数癌症。

但到了2026年8月,阿莫迪在X上的一则发帖,则多了一层现实主义色彩。阿莫迪直言,事到如今,“AI能治愈癌症”这个说法,与其说鼓舞人心,不如说是老生常谈, 而且在大多数人听来,更像是一种欺骗。

“真正有效的方法是,真正治愈癌症。”阿莫迪说道。

注:文/李馨婷,文章来源:投中网(公众号ID:China-Venture),本文为作者独立观点,不代表亿邦动力立场。

文章来源:投中网

广告
微信
朋友圈

FAQ回顾

新生实验室(Anew Labs)是什么公司?主要做什么?

新生实验室是字节跳动分拆出的AI制药公司,前身为字节跳动Seed旗下AI for Science药物发现团队,专注用AI推动生物分子结构预测、蛋白质设计等药物研发工作。2026年6月独立运营后完成2.9亿美元首轮融资,投后估值约100亿元,字节跳动保留56%股份并保持控股。

字节跳动为什么要拆分AI制药业务?

医药研发遵循“双十定律”,周期长达十年、投入十亿美元,验证周期以年为单位,且实验体系、管线推进、临床风险和商业化逻辑与字节核心业务差异大。分拆独立可降低探索成本,同时保留控股权,让团队按行业规律推进,因此字节跳动选择拆分新生实验室。

AI制药赛道为什么吸引大量资本?

AI介入后,新药研发从早期药物发现到确定临床前候选药物的周期由传统的4-5年压缩至8-12个月,节省成本且具备巨大社会价值。因此资本争相押注:国内今年AI制药融资超157亿元,超过此前3年总和;红杉中国、IDG资本、高瓴创投等头部机构都在布局。

AI制药目前真正落地的进展如何?

大部分AI制药项目仍处于临床前阶段。据《自然综述:药物发现》文章,目前仅有数十个项目处于I期和II期,III期项目寥寥无几;临床疗效才是最关键的评定维度,AI的影响尚未在后续阶段得到充分验证。

“AI能治愈癌症”的说法靠谱吗?

目前尚难兑现。OpenAI创始人奥特曼曾提出“拥有10吉瓦算力,AI或许就能找到治愈癌症的方法”,但Anthropic创始人阿莫迪在2026年指出,这个说法听起来更像欺骗,真正有效的方法就是真正治愈癌症。AI制药仍需跨过临床验证门槛。

这么好看,分享一下?

朋友圈 分享

APP内打开

赞 +1
+1
微信好友 朋友圈 新浪微博 QQ空间
关闭
收藏成功
发送
/140 0