Venture capital is quietly changing its operating model. While most funds still follow the traditional path — evaluate a pitch deck, write a check, hope for an exit — Eclipse Ventures has closed $1.3 billion with a fundamentally different ambition: to build physical AI companies from the ground up, not just invest in ones that already exist.
The firm announced the close of two new funds in early April 2026: Fund VI at $720 million and Early Growth Fund III at $591 million, bringing Eclipse’s total assets under management to approximately $10 billion. The new capital targets what Eclipse calls “physical AI” — the convergence of artificial intelligence with robotics, autonomous systems, and industrial hardware.
The Physical AI Thesis
The term “physical AI” has been circulating in venture circles for roughly 18 months, but Eclipse’s investment in the concept is one of the largest single bets on the category to date. The thesis is straightforward: most AI capital so far has flowed into software, foundation models, and cloud infrastructure. The physical world — manufacturing floors, logistics networks, power grids, construction sites, defense systems — remains underserved.
Eclipse’s target sectors span transportation, energy, infrastructure, compute, and defense. The firm has backed electric boat developer Arc, battery recycling company Redwood Materials, autonomous construction vehicle startup Bedrock Robotics, autonomous vehicle technology company Wayve, and industrial robotics lab Mind Robotics.
What distinguishes this fund is the stated intention to use a portion of the capital to incubate startups from scratch rather than purely deploying into existing companies. According to TechCrunch, Eclipse plans to build an interconnected ecosystem of startups in overlapping fields that are designed to become commercial partners as they scale — a model closer to a venture studio than a traditional VC.
Building Versus Backing
The venture-building model carries meaningful risks. Incubating a hardware or robotics startup requires longer time horizons, deeper operational involvement, and tolerance for technical failure modes that pure software investments rarely encounter. Eclipse’s broader $10 billion AUM provides the runway for that kind of patience, but it also raises the stakes: mis-built early-stage companies don’t just lose a check; they absorb years of firm attention.
Eclipse CEO Lior Susan told Bloomberg the current environment represents “the best time to build” in the firm’s history, citing declining costs for sensors, compute, and fabrication alongside increasing enterprise demand for automation driven by labor shortages and supply chain pressure.
The firm’s Cerebras connection — Eclipse was an early backer of AI chip designer Cerebras, which completed its IPO process in early 2026 — gives it direct exposure to the compute layer underpinning physical AI systems, a vertical integration that few VC firms can claim.
A Structural Shift in Deep Tech Investment
Eclipse’s $1.3 billion close arrives alongside several other signals that physical AI is becoming a fundable category in its own right. Forrester’s 2026 Emerging Technology report listed physical AI as one of its top-10 technologies, and investment in autonomous systems startups reached record levels in Q1 2026.
For founders building at the intersection of robotics, hardware, and AI, the message from Eclipse is clear: institutional capital is ready to go beyond writing checks and into the harder work of building the companies themselves. Whether that produces better outcomes than the check-only model remains to be seen over a decade-long horizon.
What is already evident is that the appetite for transforming the physical economy with AI has finally found venture capital prepared to match it in both capital and operational commitment.
| *Sources: TechCrunch — Eclipse $1.3B fund | Bloomberg — Eclipse raises for robotics, AI infrastructure | The Meridiem — venture building model* |
