Eka Robotics Raises $150M at $1.45B for Vision-Force-Action Model
By Ali Al-Montaser — Senior AI Technical Editor
Eka Robotics closed a $150M Series A led by a16z and F-Prime at a $1.45B post-money valuation. Its Vision-Force-Action model prioritizes force and touch over vision-only VLA approaches, with a force-sensitive gripper under $1,000. The robot picks 1,500 objects per hour at over 95% accuracy, versus 600 for humans.
Executive Overview
Eka Robotics, developer of the Vision-Force-Action physical AI model, has closed a $150 million Series A at a $1.45 billion post-money valuation, led by a16z and F-Prime. The company's robot picks 1,500 objects per hour at over 95% accuracy, and its force-sensitive gripper costs under $1,000. Eka's architecture departs from mainstream Vision-Language-Action (VLA) models by prioritizing force and tactile sensing to solve the last-millimeter grasping problem in robotic manipulation.
📊 Official Technical Specifications & Data Sheet
| Technical Axis | Confirmed Official Data |
|---|---|
| 💰 Pricing & Usage Cost | Eka robotic gripper: under $1,000. Series A round: $150 million. Post-money valuation: $1.45 billion. |
| 🌐 Platforms & Immediate Availability | Industrial deployment via confidential pilots with industrial customers in pick-and-place and wire-handling tasks. Not yet available as a consumer product or public API. |
| ⚡ Performance & Speed Benchmarks | 1,500 objects/hour (vs. 600 for humans), picking accuracy exceeding 95%, payloads up to 1 kilogram, generating thousands of hours of robot experience in minutes via simulation. |
| 🛡️ Security & Durability | Gripper rated for 5 million duty cycles (target: 20 million cycles, the industrial standard). Precise force control prevents crushing fragile objects (e.g., berries). |
| 🧠 Context Window | Not applicable — a physical AI model relying on force, vision, and language inputs, not a language model with a textual context window. |
| 🌍 Arabic Language & Regional Support | No official announcement of Arabic support or availability in the Arab region. Current deployment is industrial in U.S./Western markets. |
Deep-Dive Features & Architecture
Eka's team built a handling architecture fundamentally different from prevailing Vision-Language-Action (VLA) models. The Vision-Force-Action model prioritizes force and tactile inputs; co-founder Pulkit Agrawal explains that "vision tells you roughly where the object is and its pose, but the actual grasp and the actual search happen through forces." This approach addresses the last-millimeter gap that current robots struggle with, and validates criticisms from experts like Rodney Brooks that generative robotic AI models will fail without force and touch representation.
The company designed a unique robotic hand more flexible than conventional grippers, equipped with a greater number of force sensors that react to contact, which explains why it does not crush fragile berries. Technically, Eka relies on a physics simulator to model multiple forces during grasping attempts and builds a scalable reinforcement learning environment, enabling the generation of thousands of hours of robot experience in minutes. This approach moves beyond reliance on human demonstration data or worker videos and allows the robot to surpass any training demonstration provided to it, according to Professor Kris Hauser of the University of Illinois.
- Force-first control: Prioritizes tactile and force feedback over vision-only planning.
- Scalable simulation: Physics-based reinforcement learning generates thousands of hours of experience in minutes.
- Low-cost hardware: Sub-$1,000 gripper with 5 million duty-cycle rating.
- High throughput: 1,500 objects per hour at over 95% accuracy.
Benchmark & Competitive Performance
On quantitative performance, Eka's robot outperforms at 1,500 objects per hour versus roughly 600 objects for a human, an increase of approximately 150%, with accuracy exceeding 95%. By contrast, general models like Astra (an advanced ChatGPT language model) can operate a robot in an unfamiliar environment but slowly and at high cost. The fundamental difference is that Eka builds a bottom-up treatment of physics, while most of the field focuses on high-level planning. Eka's gripper at under $1,000 and rated for 5 million cycles represents a competitive price point compared with standard industrial arms targeting 20 million cycles.
Industry Impact & Enterprise Adoption
Eka's announcement carries no indication of Arabic language support or service availability in the Arab region, as current deployment focuses on confidential industrial pilots in Western markets. For developers and startups in the Arab world, the model represents an opportunity to explore integrating language planning models (such as Arabic multimodal models) with physical force-based control layers. The company's focus on force and touch could accelerate adoption in manufacturing, logistics, and agriculture, where fragile-object handling and high-throughput picking are critical. The $150 million Series A and $1.45 billion valuation signal strong investor confidence in physical AI approaches that go beyond vision-only VLA architectures.
Conclusion
Eka Robotics' $150 million Series A at a $1.45 billion valuation underscores a growing recognition that physical AI requires force and tactile sensing, not just vision and language. With a sub-$1,000 gripper, 1,500 objects per hour at over 95% accuracy, and a scalable simulation-driven reinforcement learning approach, Eka is positioning itself at the forefront of robotic manipulation. While Arabic support and regional availability remain unannounced, the company's technical direction and funding momentum make it a key player to watch in the physical AI landscape.
Media Source: TechCrunch AI | Fact Verification & Analysis: AI Tools Oasis
Frequently Asked Questions
Eka Robotics reached a post-money valuation of $1.45 billion after closing a $150 million Series A led by a16z and F-Prime.
The robotic gripper costs under $1,000 and is rated for 5 million duty cycles, with a future target of 20 million cycles, the industrial standard for standard robotic arms.
The robot picks approximately 1,500 objects per hour, each weighing up to 1 kilogram, versus about 600 objects for a human, with reliable picking accuracy exceeding 95%.
Vision-Force-Action focuses on the relationship between language, force inputs, and actions, while VLA focuses on language, visual inputs, and actions. Eka relies on force and touch to solve the last-millimeter grasping problem instead of relying on vision alone.
It was founded by MIT professor Pulkit Agrawal and former Google DeepMind researcher Tuomas Haarnoja, both leading researchers in reinforcement learning for robotic control, who first met during graduate studies at UC Berkeley.
AI Tools Oasis Team
Bringing you the latest news and analysis in the world of Artificial Intelligence with accuracy and credibility. Follow us for all updates.