While every other venture firm on Sand Hill Road has been throwing money at the latest LLM wrapper, Eclipse Ventures just closed $1.3 billion across two new funds to bet on something different: AI that touches the physical world. Robots. Chips. Autonomous vehicles. Battery recycling. Construction equipment that drives itself.

The Palo Alto firm, founded by Lior Susan in 2015, split the raise between a $591 million early-stage fund and a larger growth fund. Total assets under management now sit at roughly $10 billion. This is Eclipse's biggest raise ever, beating the $1.23 billion it closed in 2023.

The portfolio reads like a who's who of the physical AI economy: Wayve, the UK autonomous driving company backed by Microsoft. Cerebras, the AI chipmaker preparing to IPO at a $22 to $25 billion valuation. Redwood Materials, which recycles battery metals. Bedrock Robotics, building self-driving construction vehicles. Mind Robotics, an industrial robotics lab. Arc, making electric boats for commercial and defense use.

Here is the thesis in one sentence: software AI has eaten the easy problems. The hard ones live in atoms, not bits.

Eclipse has been making this argument for a decade while the rest of venture capital chased SaaS margins. Now the world is catching up. China shipped its 10,000th humanoid robot this month while Tesla has shipped zero. Japan is deploying robots in factories to solve a labor shortage that no amount of software can fix. The Pentagon is funding autonomous warships. FANUC and Nvidia just partnered to bring physical AI to industrial manufacturing.

The geopolitical tailwinds are real. Supply chain reshoring, domestic manufacturing incentives, and export controls on Chinese technology are all pushing capital toward companies that build things in the physical world. Eclipse's portfolio is positioned squarely at this intersection.

The contrarian bet is this: the biggest returns in AI over the next decade will not come from the company with the smartest chatbot. They will come from the company that figures out how to make AI work in the real world. Where things are heavy, messy, and do not follow a prompt. Eclipse just put $1.3 billion behind that conviction. Given what China is building while Silicon Valley argues about token pricing, it might be the smartest money in AI.