AMD Acquires Physical AI Startup Led by Svetlana Lazebnik for Twelve Billion Dollars
AMD has acquired a prominent AI startup focused on physical AI for approximately 1.3 trillion yen to accelerate its capabilities in robotics and real-time edge AI. This strategic move signals AMD's transition from providing pure data center silicon to delivering integrated end-to-end hardware and software platforms.
AMD has finalized the acquisition of a United States-based AI startup led by industry expert Svetlana Lazebnik in a deal valued at approximately 1.3 trillion yen. Through this acquisition, AMD aims to integrate advanced physical AI capabilities into its product portfolio. Physical AI technology enables robots, drones, and edge devices to perceive, simulate, and navigate the physical world in real time. Historically, AMD has prioritized expanding its data center accelerators to challenge Nvidia's dominance in raw compute power. This acquisition shifts AMD's focus toward vertical integration by combining chip architecture with physical control software. Rather than competing solely on silicon benchmarks, the company now targets end-to-end systems capable of running complex real-world physical simulations. While AMD has not yet released a detailed product roadmap post-acquisition, the scale of this transaction suggests deep integration into upcoming hardware. Engineers can expect advanced physics libraries and control algorithms to be incorporated directly into AMD's embedded system-on-chips and Ryzen AI processor lineups in the future.
Primary source
Open 日本経済新聞 ↗This article summarizes the source. Check the original for full details and conditions.
Comparison
| Aspect | Before / Alternative | After / This |
|---|---|---|
| Core Focus | Data center silicon performance and compute scalability | Vertical integration of silicon with physical AI and robotics software |
| Software Ecosystem | Standard ROCm drivers and runtime execution environments | Integrated physical simulation libraries and real-time edge control |
| Target Applications | Large language model training and traditional cloud workloads | Autonomous robotics, drone control, and physical edge intelligence |

