Silicon Valley Shakeup: The Secret Robotics War Between Anthropic and OpenAI

Acquisition season in Silicon Valley is moving at a punishing pace this year. Both Anthropic and OpenAI are spending hundreds of millions of dollars buying up smaller startups. They are snapping up developer tools, engineering teams, and testing platforms as fast as possible. The goal is simple: turn software models into actual enterprise revenue before the other side catches up.

Over the weekend, a massive rumor swept through tech circles on X and LinkedIn. Reports claimed that Anthropic was finalizing a deal to acquire Physical Intelligence, one of the most valuable robotics software startups in the world. Tech blogger Robert Scoble posted about the potential sale, setting off widespread speculation across the industry.

The executive team at Physical Intelligence moved quickly to damp down the chatter. Chief Financial Officer Karol Hausman sent a message to employees on Slack featuring an animated clip of a character shaking her head no, denying that a sale was taking place. Co-founder Lachy Groom stayed silent, refusing to answer public questions about the acquisition reports.

Despite those public denials, insiders confirmed that Anthropic and Physical Intelligence actually held direct acquisition talks earlier this spring. The weekend rumors might have misjudged the exact timeline, but the underlying interest between the two companies was very real.

Physical Intelligence is not some obscure research lab working on minor prototypes. Lachy Groom co-founded the company alongside top artificial intelligence researchers and professors from Google, Stanford, and Berkeley. The startup raised more than $1 billion in funding, reaching a massive $11 billion valuation in a very short time. Its core software model, known as pi0.5, operates as a primary operating system for advanced robotics research across academia and private industry.

For Anthropic, buying Physical Intelligence would instantly change its place in the market. Up to this point, Anthropic has focused almost entirely on pure software models and text processing. Instead of building an expensive hardware lab from scratch, Anthropic tested robotics through software trials. Last November, its Project Fetch initiative tested how its Claude model could control a quadruped robot dog to complete physical tasks. A follow-up test in June showed an updated model executing physical tasks 20 times faster than human operators using Claude.

While those software tests yielded impressive data, they do not replace dedicated hardware engineering. Buying an established team like Physical Intelligence allows Anthropic to skip years of tedious trial and error, jumping straight into physical hardware integration.

AI leaders increasingly realize that training software on internet text has natural limits. To build superintelligent systems, models must interact directly with the physical world. They need to process gravity, spatial awareness, force control, and real-time movement in actual physical spaces.

OpenAI realized this reality years ago. The company built an early robotics division that famously taught a robotic hand to solve a Rubik’s cube. OpenAI temporarily shut down that hardware group in 2021, with co-founder Wojciech Zaremba admitting that the industry lacked key software pieces to make physical robots smart enough. But OpenAI restarted its hardware ambitions this year, launching a dedicated robotics team under Chief Executive Officer Sam Altman to build humanoid systems and physical infrastructure tools.

The fight over Physical Intelligence gets even more complicated when you look at its cap table. OpenAI is a direct corporate investor in Physical Intelligence. Major OpenAI venture partners like Founders Fund, Khosla Ventures, and Thrive Capital also put significant money into the startup during its latest funding rounds.

OpenAI presence as a key investor gives it huge tactical advantages. Early investment deals usually include strict legal protections, such as rights of first refusal or clauses that let existing investors block a sale to a direct rival. OpenAI could potentially use those exact contractual terms to prevent Anthropic from buying Physical Intelligence.

OpenAI might even launch its own takeover offer for the startup. With both Anthropic and OpenAI filing confidential paperwork for public stock debuts this summer, securing control of top robotics talent could prove decisive. Whichever company masters physical world intelligence will hold a massive lead in the next decade of tech dominance.