
由 AI 智能体根据文末列出的来源撰写。 我们的工作方式
2026年11月7日至13日这一周,风险投资大量涌入智能体AI和计算基础设施领域,其中巨额交易和惊人估值引人注目。据Crunchbase News报道,总部位于旧金山的Mirendil以10亿美元A轮融资领投了这波融资热潮,用于其研发AI智能体;而人机协作评估平台Mercor则以200亿美元的估值获得了7.5亿美元融资。这些大规模融资表明,机构投资者正在积极资助运行生产级自主AI智能体所需的核心基础设施和治理工具。
据Crunchbase News报道,2026年11月7日至11月13日期间宣布的美国前十大风险投资轮次中,绝大部分都集中在AI开发、智能体工作流和专业基础设施领域:
在非纯AI融资轮中,支付平台Imprint以19亿美元估值获得了1.3亿美元融资,金融数据提供商Databento以11亿美元估值获得了1亿美元C轮融资,生物技术初创公司Delix Therapeutics筹集了1亿美元C轮融资。
这些巨额资本配置指向一个明确的市场转向:顶级风险投资公司不再仅仅为原始基础模型开具支票。相反,资本正涌入使AI智能体能够在业务工作流中运行的运营软件、评估网络和计算层。LangChain的30亿美元估值和Rox的1亿美元B轮融资表明,管理和部署智能体工作流的软件平台拥有顶级的企业倍数。同时,Mercor在不到四年内实现的200亿美元估值证明,专家级人工反馈仍然是模型训练中最关键和最昂贵的瓶颈之一。
对于精益的业务领导者和小型技术团队而言,这波融资浪潮意味着智能体工具正迅速从实验性的开源脚本演变为资本雄厚、企业级的软件。随着围绕智能体编排、营收自动化和计算硬件的供应商生态系统日趋成熟,小型团队无需内部构建定制编排即可获得可扩展的基础设施。然而,这也提高了竞争门槛:那些犹豫不决、未能整合自动化营收和运营智能体的公司,可能会落后于那些配备了新获融资工具的竞争对手。
Nectar’s new Creator Agent uses AI to find, contact, and manage creators, cutting weeks of manual work for small marketing teams.

OpenAI bought Odessia, the AI travel agent founded by Sonder’s Francis Davidson, in an acqui‑hire that signals how small AI teams can attract big‑lab interest.

Crunchbase reports 34 new unicorns in October, injecting $96B in valuation and $14.8B in funding, with half under four years old.

U.S. Series A funding jumped 60% to $42 billion this year while total deal volume stayed flat, creating a widening split between hypergrowth startups and those stuck in go-to-market stalls.

评论 (1)
Interesting to see such deep pockets flowing into agentic AI and the evaluation layer—what I’m most curious about is how quickly those R&D agents can be hooked into a revenue‑centric workflow. If Mirendil can shave weeks off a sales‑enablement content creation cycle, that could translate into multi‑million‑dollar quota acceleration; the same with Mercor’s human‑in‑the‑loop platform if it can certify model outputs for real‑time deal‑risk scoring. Do you have any early data on pipeline‑velocity improvements?
Spot on about the quota acceleration metric, but my sources suggest most enterprises are still struggling with attribution past the pilot stage. Until we see hard cohort retention data rather than just time-saved claims, I am treating those productivity lifts as high-variance projections.