Summary: Pittsburgh robotics startup Skild AI announced on January 14, 2026 a $1.4 billion Series C led by SoftBank’s Vision Fund 2, a price tag that puts the three-year-old company above $14 billion. Should that layer take hold, hardware manufacturers flip from competitors into prospective buyers. Measured against the roughly $500 million round at a $4.5 billion valuation it closed last summer, that is close to a triple in about seven months, and Bloomberg, citing CEO Deepak Pathak, puts cumulative funding past $2 billion.
Pittsburgh robotics startup Skild AI announced on January 14, 2026 a $1.4 billion Series C led by SoftBank’s Vision Fund 2, a price tag that puts the three-year-old company above $14 billion. Measured against the roughly $500 million round at a $4.5 billion valuation it closed last summer, that is close to a triple in about seven months, and Bloomberg, citing CEO Deepak Pathak, puts cumulative funding past $2 billion. Macquarie Capital, Nvidia’s venture arm NVentures, and Jeff Bezos — the last participating through Bezos Expeditions — also climbed aboard.
What the check buys is a narrow conviction: in robotics, the lasting value sits in software rather than in any given machine. Skild’s entire product is one family of foundation models, the “Skild Brain,” intended to run every kind of robot there is — quadrupeds, humanoids, mobile manipulators, tabletop arms. Should that layer take hold, hardware manufacturers flip from competitors into prospective buyers.
The $1.4 Billion Check That Tripled the Price Tag
Vision Fund 2 takes the lead. Bezos Expeditions, Macquarie Capital, and NVentures supplied the new money, while returning backers Coatue, Lightspeed, Sequoia, and Felicis each put in again. The strategic roster carries the stronger signal: Salesforce Ventures, LG, Schneider Electric, and CommonSpirit Health bought in, as did IQT, the US intelligence community’s venture arm. Defense, hospitals, industrial automation, appliances — any sector expecting robots at scale wants its seat at the software layer before the standard is written.
The arithmetic needs no dressing up: $4.5 billion or so in summer 2025, past $14 billion seven months on, sitting on revenue Bloomberg sizes near $30 million. Anyone who watched the 2023–2024 large language model race has seen this pattern — land-grab pricing, where capital storms the layer likeliest to set the standard and leaves commercialization for later.
Skild Brain: A Single Brain for Every Body
The company came together in 2023 in Pittsburgh around two Carnegie Mellon professors, Deepak Pathak (now chief executive) and Abhinav Gupta. At its core the pitch reads: robotics will be decided not by who gets hardware out the door first, but by whose model lands on the widest range of machines.
Officially, it is an “omni-bodied,” agentic foundation model — a single model with agent-style autonomy, running quadrupeds, humanoids, mobile manipulators, and tabletop arms. That turns decades of industrial robot practice on its head: each motion on each production line used to be hand-coded for one task and died when the task changed. The company’s own comparison is to general-purpose language models — absorb varied data once, then run wherever needed.
How the Flywheel Spins: Data Sources and Early Revenue
Data decides foundation-model winners, and the announcement names four streams: internet videos by the billion, synthetic experience in the trillions from simulation environments, teleoperation sessions, and live deployments across construction, delivery, warehouses, factory assembly, security patrol, and data centers. Every deployment trains the next model generation, which in turn wins Skild further deployments — the same compounding loop that powered the language-model takeoff.
Revenue, unusually for robotics, showed up early: Bloomberg reports about $30 million within months of first sales. The Series C money goes where you would expect — growing the data sources, financing foundational research, pushing real-world deployments wider.
The Timing Question: The Contest for the Robot’s Operating Layer Is Already Crowded
Skild has plenty of company. Field AI sells robot software built to adapt quickly across different bodies, and 1X has just put out a world model so its humanoids can train inside simulation first, then meet the physical world. One level up, CES 2026 made “physical AI” the theme of the whole show (our recap) — hardware demos arrive every year; what everyone keeps calling missing is an operating layer that spans machines. SoftBank, with several robotics-adjacent positions on its books, plainly reads this window as open.
Takeaways for Teams Shipping Products
Three lessons to carry home. First, should the model layer turn robot hardware into a commodity, the differentiation fight moves to workflows and integrations — exactly where value went once smartphone hardware became standardized. Second, the data recipe of simulation first and video second is now table stakes; teams building embodied or software agents can adopt the approach wholesale. Third, a $14 billion valuation standing on only about $30 million in revenue tells you the market is underwriting first-mover positioning rather than the numbers — expect the race settled across the next one or two fundraising rounds, and time vendor commitments to match.