Exclusive: Google deepens Thinking Machines Lab ties with new multi-billion-dollar deal
Summary (EN)
TechCrunch reported that Thinking Machines Lab, the startup founded by former OpenAI CTO Mira Murati, has signed a new multi-billion-dollar agreement to expand its use of Google Cloud infrastructure. According to the report, the deal is worth a single-digit number of billions of dollars and includes access to systems powered by Nvidia’s latest GB300 chips, along with cloud services needed for large-scale model training and deployment. The report says this is Thinking Machines’ first deal with a cloud provider, even though the arrangement is not exclusive and the company may still work with multiple infrastructure partners over time. Google is described as using the agreement to strengthen its position with frontier model builders by bundling compute with storage, Kubernetes, database services, and broader cloud infrastructure. The story also offers a clue to Thinking Machines’ technical direction. Google’s own release reportedly highlighted support for reinforcement learning workloads, suggesting that the startup’s product stack, including its previously launched Tinker system for automating custom frontier model creation, depends heavily on RL-style training and serving. The importance of this development lies in what it says about industry structure: cloud providers are no longer merely vendors of commodity compute, but strategic counterparts competing to lock in elite frontier labs early. In the current market, access to advanced GPU systems and integrated deployment infrastructure has become a bargaining lever that can shape where and how major model companies build, making this agreement a notable signal about capital intensity, supply competition, and platform alignment in frontier AI.
Summary (ZH)
TechCrunch 独家报道, 由前 OpenAI CTO Mira Murati 创立的 Thinking Machines Lab 已与 Google Cloud 签署新的多十亿美元级合作协议, 以扩大其对 Google 云端 AI 基础设施的使用。报道指出, 该协议价值为数十亿美元中的个位数区间, 内容包括使用搭载 Nvidia 最新 GB300 芯片的系统, 以及支撑大规模模型训练和部署所需的云服务能力。这被视为 Thinking Machines 首次与云服务商达成正式合作, 尽管协议并非排他, 公司未来仍可能同时采用多个基础设施伙伴。对 Google 来说, 这笔合作体现了其争夺前沿模型实验室的策略, 即把算力、存储、Kubernetes 引擎、数据库和整体云栈打包为一体, 争取在早期绑定高速增长的模型公司。报道还透露了 Thinking Machines 的技术方向。Google 在相关表述中重点提到强化学习工作负载支持, 说明该公司此前推出的 Tinker 等产品架构很可能高度依赖强化学习式训练与部署流程。这一事件的价值不只在于交易金额, 更在于它揭示出 AI 产业结构正在变化, 云厂商不再只是通用算力供应商, 而是通过先进 GPU、部署平台和服务整合能力, 成为前沿模型公司在技术路线和资源配置上的关键战略伙伴。
Source