Artificial intelligence companies are moving beyond software into the physical world, and a recent rumor around Anthropic and robotics startup Physical Intelligence captured that shift. The speculation spread quickly online, even after the startup's CEO denied that an acquisition was happening.
The attention was not surprising. Physical Intelligence has become one of the most closely watched names in robotics, backed by major investors and known for its π0.5 model, which is widely discussed in robotics research circles. The company was founded by Lachy Groom, former Google researchers, and academics from Stanford and Berkeley, and has already raised more than $1 billion.
According to reporting from The Information, Anthropic and Physical Intelligence did hold acquisition talks this spring. That detail suggests the rumor had a real foundation, even if the public version of the story was inaccurate. Physical Intelligence CEO Karol Hausman reportedly told employees the reports were untrue in a Slack message.
The broader context is just as important. Anthropic has made several acquisitions this year, while OpenAI has moved even faster, buying a long list of startups since 2023 as both companies prepare for potential public listings. Their parallel expansion shows how quickly the AI race is evolving from model development to product ecosystems and enterprise scale.
Robotics is emerging as a strategic frontier because real-world interaction adds a new layer of intelligence that text alone cannot provide. OpenAI has already returned to robotics with a renewed lab effort, while Anthropic has explored the field through research projects such as Project Fetch, which tested how Claude could help people program a robot dog.
Physical Intelligence also sits at a notable intersection of investors and competitors, with OpenAI among its backers. That makes any future move around the company especially significant in the broader AI landscape. As AI firms deepen their robotics ambitions, the next wave of innovation may be shaped by how well machines learn from the physical world.
