On January 14, 2026, Pittsburgh-based Skild AI announced a $1.4 billion Series C led by SoftBank’s Vision Fund 2, valuing the three-year-old startup at more than $14 billion. That is roughly triple the $4.5 billion valuation attached to the roughly $500 million round it closed last summer, and it lifts Skild’s total funding above $2 billion, CEO Deepak Pathak told Bloomberg. Nvidia’s venture arm NVentures, Macquarie Capital, and Jeff Bezos through Bezos Expeditions all joined the round.
The bet behind the check is specific: in robotics, the durable value sits in software, not in any particular machine. Skild builds one set of foundation models — the “Skild Brain” — meant to power any robot: humanoids, quadrupeds, tabletop arms, mobile manipulators. If that layer works, every hardware maker stops being a rival and starts being a customer.
A $1.4 Billion Round That Tripled the Valuation
SoftBank led through Vision Fund 2. Alongside NVentures, Macquarie Capital, and Bezos Expeditions, earlier backers Lightspeed, Felicis, Coatue, and Sequoia all doubled down. The strategic names are the more interesting signal: LG, Schneider Electric, CommonSpirit Health, and Salesforce Ventures took positions, as did IQT, the venture arm of the US intelligence community. Appliances, industrial automation, hospitals, defense — every industry that expects to run robots at scale wants a seat at the software layer before the standard gets set.
The valuation math is blunt: about $4.5 billion in summer 2025, more than $14 billion seven months later, on revenue that Bloomberg pegs at roughly $30 million. That is land-grab pricing, familiar to anyone who watched the 2023–2024 large language model race — capital rushes the layer that could become the standard, then worries about commercialization later.
Skild Brain: One Model, Any Robot
Skild AI was founded in 2023 in Pittsburgh by two Carnegie Mellon professors, Deepak Pathak (now CEO) and Abhinav Gupta. Its premise, stated plainly, is that the robotics story is not about whose hardware ships first but about whose model can deploy across the most kinds of machines.
The product is an “omni-bodied,” agentic foundation model: one model, with agent-style task execution, controlling quadrupeds, humanoids, tabletop arms, and mobile manipulators. It is the opposite of how industrial robots have been built for decades, where every motion on every line is task-specific code that dies when the task changes. The explicit analogy is to general-purpose language models: learn from diverse data, deploy anywhere.
The Data Flywheel and Early Revenue
Foundation models are won on data, and Skild’s announcement lists four sources: simulation environments generating trillions of synthetic experiences, billions of internet videos, teleoperation sessions, and live deployments across security patrol, construction, delivery, data centers, warehouses, and factory assembly. Deployments feed the next generation of the model, which qualifies Skild for more deployments — the same compounding loop that made language models accelerate.
Commercial traction arrived faster than is typical for robotics. Bloomberg reports Skild generated about $30 million in revenue within months of starting to sell. The stated use of the Series C is what you would expect: scale the data sources, fund foundational research, and expand real-world deployments.
Why Now: A Crowded Race for the Robot’s Operating Layer
Skild is not alone. Field AI is selling similarly adaptable robot software, and 1X just released a world model so its humanoid robots can learn in simulation before touching the physical world. Further upstream, CES 2026 put “physical AI” at the center of the entire show (our recap) — hardware demos are an annual event, but the piece everyone keeps pointing at as missing is a cross-machine operating layer. SoftBank, which is placing several robotics-adjacent bets, evidently judges that this window is opening now.
What It Means for Builders
Three takeaways. First, if the model layer commoditizes robot hardware, differentiation shifts to workflows and integrations — the same migration value took after smartphone hardware standardized. Second, the simulation-first, video-supplemented data recipe is now standard practice, and any team training embodied or software agents should borrow it directly. Third, a $14 billion valuation on roughly $30 million of revenue means the market is pricing first-mover advantage, not financials — expect this race to be decided within the next one or two funding cycles, and plan vendor commitments accordingly.
Sources
- Announcing Series C — Skild AI
- Robotics software maker Skild AI hits $14B valuation — TechCrunch
- Skild AI lands $1.4B, tripling valuation to $14B — Crunchbase News
AI-assisted summary compiled from the sources above, reviewed by a human before publishing.