VERTEBRATE.ai

Eka, a startup building what it calls a foundation model for the physical world, has reached a $1.45 billion valuation, the company confirmed this week as part of a combined $380 million funding round raised jointly with Arda, another entrant in the increasingly crowded field known as physical AI. The two companies did not disclose how the $380 million splits between them, nor did they name the investors behind the round, but the headline number alone places Eka among the more richly valued startups in robotics-adjacent software, a sector that a year ago struggled to command even a fraction of that attention.

What distinguishes Eka from the parade of robotics startups chasing capital this year is less the size of the check than the source of the enthusiasm. Researchers who have examined Eka's underlying model have reportedly described it in terms usually reserved for breakthroughs in large language models — not because it writes better prose, but because it appears to reason more capably about the physical world: how objects move, how forces interact, how a robotic arm might fail before it fails. That is the holy grail of "physical AI," a term that has become shorthand for systems meant to give machines something closer to common-sense intuition about bodies, gravity, and friction, rather than just the pattern-matching that powers chatbots.

The joint raise with Arda, and the fact that both companies emerged from the same funding cycle with blockbuster figures, says as much about the market's mood as it does about either company's technology. Investors have spent the past two years pouring money into humanoid robots and warehouse automation, often rewarding spectacle — a robot that can fold laundry or walk a catwalk — over the less photogenic work of building the models that let those machines understand their environment in the first place. Eka's valuation suggests a shift in where the smart money is willing to bet: not on the hardware shell, but on the cognitive layer underneath it, the software that might eventually be licensed across dozens of robot manufacturers the way operating systems once were across PC makers.

That distinction matters beyond Silicon Valley term sheets. If physical AI models mature the way large language models did, the companies that own the best ones could become indispensable infrastructure for an entire robotics economy — manufacturers, logistics firms, even hospitals — in the same way that a handful of AI labs now effectively set the terms for the chatbot industry. For workers in warehouses and factories, that could mean machines that adapt faster to new tasks without months of costly reprogramming. For investors, it raises the now-familiar question of whether valuations like Eka's reflect durable technological advantage or simply the latest round of a funding mania that has already produced several billion-dollar robotics bets this year alone, with no guarantee of which ones will still be standing when the hype settles.