If you’ve been focused on developing software, you may have overlooked the fact that chatbots aren’t now the most intriguing tech fundraising story. Robots are involved. In particular, investors are referring to this area as “physical AI”, AI systems that perceive, navigate, and act in the actual environment in addition to answering questions, and the amount of money pouring into it this year is quite astounding.
The numbers, briefly
In 2025, robotics and physical AI start-ups raised a record amount, more than double the previous year, and 2026 is expected to surpass that. According to one tracker, worldwide robotics financing reached $18.8 billion in just the first half of this year, well beyond both the previous funding peak established back in 2021 and the full-year figure for 2025.
Deal value increased significantly even as the number of individual agreements decreased, according to a different report on the second quarter. This indicates that while fewer companies are receiving funding, the cheques given to those that do are growing significantly.
The CEO of Nvidia allegedly said that the “ChatGPT moment for robotics” has arrived at CES back in January. Of course, that’s a marketing slogan, but the money behind it isn’t.
Where the money is actually concentrating
The headline total is not as important as this section. The money is pouring into a small number of layers rather than being distributed equally throughout “robots” as a category. This year, the vast majority of revealed capital in the field has gone to robotic foundation models and general-purpose robots, with humanoids and other specialised categories fighting for a much smaller residual portion.
Additionally, the dollar total that remains after removing the few biggest rounds is a tiny portion of the headline figure; this is a mega-round-led market rather than a wide, equally spread one.
Another indication worth keeping an eye on is that the great majority of agreements and funds this year are flowing into follow-on rounds for businesses that already have some momentum rather than brand-new, first-time wagers. Instead of funding the category from the ground up, investors are instead supporting teams who already have technology, a customer channel, or a strategic partner.
The creation of new businesses in physical AI hasn’t halted, but it’s happening much more covertly and mostly at the infrastructure and tools layer rather than another capital-intensive humanoid startup attempting to compete directly.
Why this should matter even if you’re not building a robot
This may seem like a tale about someone else’s industry if you work in software development. For two reasons, it isn’t.
First, the pattern is a preview of where AI capital discipline is heading generally. The behaviour here is instructive: concentrated bets on proven teams, follow-on financing over first checks, and genuine scepticism toward “us too” firms entering a crowded tier of the stack. The same investors funding physical AI are also the ones selecting whether to fund your next round. Expect the same standard if you’re raising in any AI-related category: proof of deployment, not just an engaging architectural narrative.
Second, many software companies that do not consider themselves robotics companies are discreetly turning to physical AI as infrastructure. This year, warehouse and logistics automation, industrial inspection, autonomous delivery, and related fields are transitioning from pilot to actual deployment; these are large-scale, real-world experiments rather than demonstrations. Whether or not you are the one creating the robot, your clients’ perceptions of what is automatable are currently changing if your product has any connection to supply chain, logistics, manufacturing, or field operations.
The specific place to pay attention: bottleneck layers, not robot bodies
If there’s one useful takeaway from the data for a founder outside of robotics, it’s this: the bar for starting a real robot company has increased significantly; you now need hardware depth, a data strategy, manufacturing access, deployment partners, and AI talent all at once, which is a lot to put together from scratch.
However, the picks-and-shovels layer, the layer beneath the robots that deals with tooling, safety, simulation, and orchestration, remains really open and is the layer that investors have identified as having the most potential for new players.
Compared to “compete with a humanoid robot company,” it is a far more accessible location for a software-first team to create, and it’s also where many of the more intriguing new company formations this year are taking place.
What to actually take from this
To learn something practical here, you don’t have to switch to robotics. Regardless of what you’re building, there are three things to consider:
• Proven team with early traction is now outperforming “compelling technology story” in all areas of AI, not just physical AI. It is worthwhile to correct your pitch before your next raise, regardless of the category, if it heavily relies on technology and lacks deployment evidence.
• If your clients are involved in manufacturing, field service, logistics, or other physical operations, you should expect that automation is advancing more quickly than your present strategy takes into consideration. It would be worthwhile to discuss what your most operationally orientated clients are currently testing.
• It is frequently more affordable to construct a hot category’s unglamorous infrastructure layer rather than its ostentatious end. This year’s physical AI reflects this, as has every previous capital-intensive tech wave.
Your job won’t be replaced by robots. However, the way one category is receiving funding is a fairly strong predictor of the tone of all AI-related fundraising discussions for the remainder of the year.
Send this to any team members who believe “physical AI” belongs to someone else. Next week: a few specific instances of what “picks and shovels” truly entails as a startup thesis in 2026.
