Robotics startup Skild AI raised $1.4 billion on January 14 in one of the industry’s largest funding rounds ever, reaching a $14 billion valuation. That’s triple its $4.5 billion valuation from just seven months ago.
SoftBank Vision Fund led the Series C, with Nvidia, Jeff Bezos, Samsung, LG, and Salesforce participating. The investor mix isn’t random—it maps directly to deployment: chip makers (Nvidia), hardware manufacturers (Samsung, LG), and enterprise software (Salesforce).
Founded in 2023 by Carnegie Mellon professors Deepak Pathak and Abhinav Gupta, Skild AI is building what they call a “universal robot brain”—one AI model that controls any robot form without custom training. The company is already deployed in data centers and factories, generating over $30 million in annual revenue.
Omni-Bodied AI: One Model, Any Robot
Traditional robotics requires task-specific programming for each robot design. ROS (Robot Operating System) developers know the drill: custom training, specific hardware configurations, rigid scripts.
Skild AI’s approach is different. They trained their model on 100,000 simulated robots with different morphologies—humanoids, quadrupeds, robotic arms, mobile manipulators. The training volume is reportedly 1,000x larger than competing models. This massive scale forces the AI to learn general principles of locomotion and balance rather than memorize sequences for specific hardware.
The result: their AI adapts via in-context learning, similar to how large language models adjust to new tasks. Deploy it on an unfamiliar robot body, and it figures out how to control it. Break a robot’s limb mid-task, and it compensates without retraining.
For developers, this changes the deployment model. No custom training phase. No robotics PhD required. Focus shifts from low-level motor control to high-level task design. If it scales, this could standardize robotics development the way Android standardized mobile—one software layer across all hardware.
Enterprise Reality Check
Most robotics startups announce funding rounds before shipping actual products. Skild AI is different—they’re already deployed.
Current use cases include data center infrastructure automation, factory floor manufacturing, warehouse logistics, security patrols, and construction tasks. The company went from zero to $30 million in annualized revenue “in just a few months” during 2025, according to investor communications.
The enterprise-first strategy is smart. Controlled environments like data centers and factories reduce variables during the scale-up phase. Consumer homes—with pets, kids, clutter, and chaos—are exponentially harder. Skild is targeting those unstructured environments long-term, but proving the technology in predictable settings first.
SoftBank’s Physical AI Bet
SoftBank isn’t just writing a check. In October 2025, they announced a $5.4 billion acquisition of ABB Robotics, set to close mid-2026. Pairing ABB’s hardware manufacturing with Skild’s universal software is the strategy.
Timing matters too. The Skild AI funding came five days after CES 2026, where “physical AI” dominated the conference. Nvidia CEO Jensen Huang called it “the ChatGPT moment for physical AI.” SoftBank CEO Masayoshi Son has been vocal about positioning the Vision Fund as a leader in robotics and AI, calling the ABB deal part of their “Physical AI” expansion.
The broader market validates the timing. The global robotics market hit $124 billion in 2026, with robot installations expected to surpass 700,000 units by 2028. Robotics software alone will generate $24.5 billion in revenue by 2030. This isn’t a speculative play—it’s infrastructure for industrial automation that’s already scaling.
The Skeptical Take
A $14 billion valuation on $30 million in revenue is a 467x revenue multiple. That’s extraordinarily high, even by growth-stage AI standards. The valuation prices in massive future adoption that hasn’t materialized yet.
There’s also the universal capability question. Can one model truly handle all robotics use cases? Simulations train AI on idealized scenarios, but real-world environments are messy. A robot that excels in a data center might fail in a cluttered warehouse or unpredictable home.
Previous robotics hype cycles—autonomous vehicles, delivery drones, home robots—took far longer to deliver on promises than investors expected. Foundation models worked for language (LLMs), but physical embodiment introduces constraints that text generation doesn’t face: physics, latency, safety, hardware limitations.
That said, the counterarguments are strong. The founders have deep robotics and AI research backgrounds. The company is generating real enterprise revenue, not burning cash on demos. Strategic investors like Nvidia, Samsung, and LG are positioning for integration, not speculation. And the 1,000x training data advantage creates a defensible moat if the technology proves out.
What to Watch
Robotics is entering the foundation model era. Whether Skild AI becomes the “Android for robotics” depends on two things: whether their omni-bodied approach generalizes beyond controlled enterprise settings, and whether the $14 billion valuation reflects realistic scaling timelines.
The enterprise deployments are real. The funding is real. The strategic investor lineup suggests near-term adoption, not distant-future bets. But high expectations create high risk. If Skild AI can deliver on universal robotics capabilities at scale, this funding round will look prescient. If not, it’ll join the long list of overhyped robotics plays that underdelivered.
The industry shift from narrow task-specific robots to general-purpose AI is happening. The question is how fast, and whether Skild AI will lead it.
