Bringing Robots to the Boring Industries with Physical Fund's Nima Ashraf
Nima Ashraf is the founder and managing director of Physical Fund, a pre-seed venture fund investing in robotics and automation startups building at the intersection of hardware and software.
Nima Ashraf is the founder and managing director of Physical Fund, a pre-seed venture fund investing in robotics and automation startups building at the intersection of hardware and software.
Can you walk me through your career and what led you to launching Physical Fund?
My background is in mechanical engineering, I always wanted to build cool products and change the world. I got my first taste of startup life at MakerBot during the early 3D printing boom through its acquisition, followed by the autonomous mobility space at Zoox which was acquired by Amazon, and eventually scaling cooking robots at Miso Robotics.
There's obviously a theme there. This trajectory taught me that robotics is the ultimate blend of hardware, software, and design. I was interested in investing early on but wanted to build things first, along the way I realized how much of a gap there was. There weren't many early-stage investors in the space supporting technical people, with cool ideas, who wanted to experiment and do crazy stuff. I wanted to be the partner I wished I had, the one who understands the unique pain of scaling robots and automation.
What are the basics of the fund in terms of thesis, check size, and target AUM?
We are a high conviction, pre-seed fund focused on robotics and automation. Physical AI is the modern term for it, but really it's about intelligence entering the real world through machines and robots. My thesis is specifically on vertically integrated platforms where intelligence interacts with the real world. The AUM target is $2 million, and check sizes are between $50,000 to $75,000. We’ve already deployed into our first few investments and are moving fast.
Physical AI is a term that gets thrown around a lot. Is it just a new name for robotics, or is there something fundamentally different about what it describes?
Both are true. Initially, I dismissed it as marketing fluff, it felt like rebranding something that people in my world have been working on for many years. But there is a technical shift happening, we’re moving from deterministic robotics to probabilistic. Modern robots can now perceive and operate in dynamic, unstructured environments that were previously impossible to automate. Delivering reliability is still the biggest challenge, but it’s no longer just about the body of the robot. It’s about the full stack intelligence that allows it to reason through the real world.
A lot of roboticists who have been in the game for decades are pretty cynical about the next big wave. This time does feel different, though. Why?
It feels different to me because the barriers to entry are lowering. On one side, AI is automating the heavy lifting of software development; on the other, hardware is beginning to commoditize into accessible, open-source components. This unlocked creativity means a mechanical engineer, like me, can now do some of the work of a software team, and vice versa. But the real shift is on the customer side, which started a few years ago. The labor crisis has made automation a survival requirement for industry, not a luxury. The core challenges of delivering automation that is valuable and reliable are the same, but the tech is finally robust enough to move at the speed of the problem.
What areas of physical AI are getting too much attention right now, and what's flying under the radar?
Humanoids in the home are the obvious one to me; they make for great viral videos but have little immediate utility or ROI. The silver lining of humanoids right now is that it’s driving public interest. The real alpha is in the boring legacy industries; manufacturing, agriculture, logistics, and general autonomy. Most mid-market companies are still operating with manual processes that are ripe for automation. I want to back the teams building the essential infrastructure for the physical world, the kind of tech that doesn't just look cool but actually works at scale.
What are the biggest bottlenecks to deploying physical AI in the real world?
The lab-to-field gap is still the ultimate bottleneck. It’s easy to make a robot work once for a demo or video, it’s fundamentally difficult to ensure deployment reliability and 99.9% uptime in a dirty or dynamic environment. Data scarcity is also a huge bottleneck. Unlike LLMs, physical AI can't just scrape the internet, it needs high quality real world data. Solving for both of those is really where the rubber hits the road.
When you're evaluating founders in this space, what are you looking for?
I like to think of it as the full-stack founder. This means someone with the technical depth to build the product, but the entrepreneurial hustle to survive and thrive even through an extended industrial sales cycle. The other big requirement is a genuinely unique insight. Usually that means someone who has lived the problem, maybe at a much larger company, and realized the existing tools were broken. It's tough to sell robots to anyone, selling into legacy industries requires trust and a serious solution. Most customers aren't looking for a toy, they're looking for an operational partner.
How do you define deep tech?
Deep tech isn't a marketing category, it’s a technical moat. I define it as any breakthrough in science or engineering that is fundamentally difficult to replicate, even with capital. It usually requires specialized knowledge, a novel engineering application, or a barrier to entry built on technical difficulty. Is it a unique solution that’s hard to replicate? That’s deep tech.
Infinite Frontiers is made possible by HAUS, the leading public relations agency for deep tech startups and scaleups.
