Every SaaS investor is claiming software is “dead” and is now heavily rotating into hard tech. Robotics, energy, defense, manufacturing and many others.
There’s genuine reason to be excited on the tech side. Robotics is finally having its moment. Self learning, models getting better, BOM cost coming down, robots actually deployed. Humanoid manufacturing costs fell about 40% in a year. One humanoid maker’s robots have logged over 65,000 hours across nine customer facilities. It’s inevitable. But deploying robots will take longer.
While doing diligence for a startup, I called up a factory where a robot is deployed, and I got to know that it was not on the production line. Even though the founder claimed it was. I called up another factory where the main point of contact was the champion. The company thought they could quickly expand to every factory in the portfolio. But when asked who gets to make that decision, the champion was very honest, saying that it is an individual decision where they can have some influence, but definitely not enough authority to make the change themselves. This is the norm, not an exception.
There’s a deployment gap, similar to how there is a diffusion gap in AI in enterprises. All the layered cake was already there for AI to diffuse, and even then, when we talk to companies, the diffusion is still missing. As of May 2026, 19.8% of US businesses used AI in any business function.
What’s happening in robotics though is a bit different: efficiency has ruled the roost in all these old school industries, especially factories and manufacturing, where they’ve seen too many false dawns. We’ve had digital twins and Industry 4.0. People have had arms that were promised: hey, we’ll move from arms that are doing picking to full-scale robotics coming in.
The companies just can’t afford to miss one shift. There’s a top down push for AI and real labor shortages. People are trying, but it’s hard to give up efficiency and years of human programming. So they are willing to try and do pilots with a lot of these companies.
In software we joke about contracted ARR. In robotics, it’s about a lot of pilots where everybody thinks those pilots will convert. If you get one factory, you will get 500 factories. But that’s not how this typically works. Carmakers are piloting single or double digit numbers of robots per plant. There are factory managers who control this. It’s not always top down. Even if there is a fancy partnership that’s announced, it’s not necessary that these robots will deploy.
It is extremely hard to get access to production lines. We almost want these companies to learn to fail, to make mistakes, so that they can then improve their models. A 16-robot fleet learning from its own work reached a 95% average success rate. Models will get there. But if you give it a sideline where real work is not happening and it’s just a showpiece, then you’re not learning as fast as possible, and that’s what’s happening in the robotics market right now.
Now sure, this time IS different, but atoms just work slower than bits, even though sheer will can move mountains. I spent 10 years working in the physical world. I deployed networks in fulfillment centers and watched how a simple riser installation could delay getting a network online for weeks. How transferring utilities could take weeks. How a simple zone 1-7 delivery could have inconsistent delivery times. Roofs don't fix overnight.
What robotics companies especially need is more patient capital. Matic went through many product iterations, and so did SpaceX, and that requires patience.
There’s a lot of impatient capital in the market. Robotics startups raised $18.8B in the first half of 2026, more than in all of 2025. My biggest worry for robotics founders taking money from software investors is that everybody’s investing, thinking they will get a return at the same IRR as software. Capital is the lifeblood, especially given high capex. It’s why founders will take it, not knowing it’s impatient.
The time to invest is now, but if we’re looking to see Anthropic-like growth, then we’re mistaken. So my push to investors (and founders) is simple: Do the diligence. Spend time in factories. Get out of that cushy chair in Silicon Valley. Do a few tours. Make friends with plant managers. Understand the problem deeply, since most of us don’t get it and it’s not happening at the coasts.
The future is glorious, but it’ll take time. For the next board meeting, try to understand these gaps and figure out how to remove bottlenecks versus simply asking to make that revenue number go up.
There is a "How".