Eka机器人公司融资1.5亿美元，估值达14.5亿美元

机器人公司Eka完成A轮融资1.5亿美元，由a16z和F-Prime领投，融资后估值达到14.5亿美元1。公司开发的物理AI模型采用视觉-力-动作架构，配合力感应夹爪，使机器人在物体拾取任务上的表现超越人类水平1。

在性能指标上，Eka机器人每小时可靠地拾取约1500个物体，相比人类每小时约600个的水平，拾取准确率超过95%1。支撑这一性能的硬件成本相对低廉，夹爪硬件成本低于1000美元1。目前硬件设计寿命为500万次工作周期，公司目标是将其提升至2000万次1。

a16z合伙人Raghu Raguram表示，将在未来几个月实现多项灵巧操作任务的超人类表现1。Eka由麻省理工学院教授Pulkit Agrawal和前谷歌DeepMind研究员Tuomas Haarnoja共同创办1。

Eka Robotics has completed a Series A funding round of $150 million, valuing the company at $1.45 billion 1. The round was led by a16z and F-Prime 1.

The company has developed a physical AI model built on a vision-force-action architecture paired with force-sensing grippers that enable robots to perform object-picking tasks at superhuman levels 1. The system can reliably pick approximately 1,500 objects per hour with an accuracy rate exceeding 95%, outperforming human operators who pick around 600 objects hourly 1. The force-sensing gripper hardware costs less than $1,000 1.

Currently, the hardware is designed for a lifespan of 5 million work cycles, with the company targeting an eventual capacity of 20 million cycles 1. The technology bridges the gap between generality and performance in scaled deployment through physical simulation and reinforcement learning 1. According to a16z partner Raghu Raguram, the company plans to demonstrate superhuman performance across multiple dexterous manipulation tasks in the coming months 1.

Eka was founded by MIT professor Pulkit Agrawal and Tuomas Haarnoja, a former researcher at Google DeepMind 1.