Latest / The Robotics Podcast with Fexingo: Autonomous Systems, Industrial Robots, and Hardware / Why Robot Software Is The Real Bottleneck
Transcript
- Lucas: I was looking at some recent deployment reports from three major logistics hubs, and the data is pretty stark. The hardware is there, the arms are precise, the sensors are cheap enough now, but the time it takes to actually get these robots working on the floor has gone up, not down. Luna: Right, because every vendor brings their own proprietary OS. You can't just plug a new arm into an existing line without rewriting half the codebase. Lucas: Exactly. It is a fragmentation issue that nobody wanted to talk about during the hype cycle. If you go to a factory floor today, you might have conveyors controlled by Siemens, AGVs from one startup, and manipulators from another. They do not speak the same language natively. Luna: So we are seeing a lot of dead capital sitting on warehouse floors while engineers try to make them talk to each other. Lucas: Precisely. And the cost of that integration labor is eating into the margins that were supposed to justify the automation in the first place. I want to drill into this idea of middleware—the software layer that tries to translate between these different hardware protocols—and why it is becoming the single most valuable piece of intellectual property in the robotics sector right now. Luna: It feels like we are repeating the early days of cloud computing before AWS standardized everything. Is that the parallel? Lucas: That is a really sharp comparison. But with a twist. In cloud computing, the underlying commodity was compute power, which became a utility. In robotics, the commodity is physical movement, which is infinitely more complex to standardize because physics does not care about your API. Luna: So you cannot just abstract away gravity and friction like you can abstract away server racks. Lucas: Correct. You have to account for wear and tear, temperature changes, and load variations. This is why companies like Boston Dynamics and Figure AI are pouring so much engineering talent into their control stacks rather than just building prettier shells. Luna: I read that Figure recently open-sourced part of their motion planning library. Was that a strategic move or just desperation? Lucas: It was a strategic move to accelerate ecosystem adoption. If they make their robot easier to program for third-party integrators, those integrators will recommend the robot. It is a classic platform play, but executed in a market that is still very hardware-centric. Luna: But does that actually solve the integration problem for the end user, or just shift the burden to the developer? Lucas: It shifts the burden, yes. But the real bottleneck is that most factories do not have developers who understand robotics. They have maintenance technicians who understand mechanical fixturing. When you force them to write Python scripts to calibrate a gripper, you create a massive training deficit. Luna: So the software needs to become invisible again. It needs to be declarative rather than imperative. Lucas: Exactly. Tell the robot what you want it to do, not how to move its joints to do it. That is the holy grail of robotic middleware. And looking at the market as of November first twenty twenty six, we are seeing two distinct approaches trying to win this race. Luna: Which two approaches are dominating the conversation right now? Lucas: One is the vertical integration play, where a company builds both the hardware and the software stack to ensure they talk perfectly. The other is the horizontal abstraction layer, where a software company sits on top of all the hardware vendors and provides a unified interface. Luna: The vertical approach seems safer for quality, but the horizontal approach scales faster if they can handle the complexity. Lucas: Scalability is the key word here. Because no single factory is going to buy robots from only one vendor forever. They will always mix and match based on price and availability. So the horizontal layer has a larger total addressable market, provided it can solve the latency issues. Luna: Latency being the enemy of safety and precision in any physical system. Lucas: Right. If your middleware adds fifty milliseconds of delay to a collision detection algorithm, you are not just slower, you are dangerous. That is why many large manufacturers are hesitant to trust third-party software with critical safety functions. Luna: So we see a lot of pilot programs that never scale past the proof of concept stage because of this trust gap. Lucas: Absolutely. And that is where the investment community is getting confused. They see the hardware sales numbers and assume growth is guaranteed. But if the software integration costs remain high, the recurring revenue model for these robotics firms looks much thinner than it appears. Luna: It reminds me of the enterprise software wars in the nineties. The winner was rarely the best product, but the easiest to deploy. Lucas: I think it is even more extreme now. In the nineties, you had IT departments to handle the deployment. Today, the people deploying these robots are often the same ones fixing the jams in the assembly line. The bar for usability has to be near zero. Luna: So we are going to see a consolidation of middleware providers. Too many small players won't survive the integration overhead. Lucas: I agree. We will likely see two or three dominant platforms emerge within the next five years. Everything else becomes a niche tool or gets acquired. Luna: Before we dig deeper into the acquisition landscape, I want to take a quick moment to talk about how we keep this show running. Lucas: Yeah, we deliberately do not run ads on these shows. If you want to support that choice and help us keep the research deep, the link is buy me a coffee dot com slash fexingo — that's F. E. X. I. N. G. O. Luna: It really makes a difference for our ability to cover these niche tech angles without compromise. Lucas: Back to the market structure. If we look at the vertical integrators, their advantage is speed. Because they control the kernel, they can optimize the feedback loop between perception and action in a way that a middleware layer simply cannot match. Luna: But they are trapped by their own hardware limitations. If their robot design fails in the market, their software becomes worthless. Lucas: That is a valid risk. Hardware cycles are long and capital intensive. A failed hardware prototype can sink a company for years. Middleware companies have lower barriers to entry, but they face constant churn as hardware vendors change their APIs. Luna: So the middleware companies have to move fast to stay relevant, almost like a SaaS company in a volatile market. Lucas: Exactly. And the interesting dynamic is that some of the biggest tech giants are quietly buying these middleware startups. Not to compete with the hardware makers, but to secure the software layer for their own broader automation ambitions. Luna: Like Google or Microsoft trying to own the brain, not the body. Lucas: That is a fair way to put it. And it raises the question of whether we will ever see a true open-source standard for robotics, similar to Linux for servers. Luna: Linux worked because the hardware was standardized early on. Robotics hardware is still wildly diverse. Lucas: True. But we might see something like ROS, the Robot Operating System, evolve into a de facto standard for communication, even if the execution remains proprietary. It is already happening in academic and research circles. Luna: Will that trickle down to industrial settings? Or will safety certifications kill open source adoption? Lucas: Certification is a huge hurdle. Industrial safety standards require deterministic behavior, which is hard to guarantee in a complex, open-source codebase. Most factories will stick to certified, closed-loop systems for critical tasks. Luna: So we are heading toward a hybrid model. Open source for experimentation and non-critical tasks, closed source for high-stakes production. Lucas: I think that is exactly where we are heading. The industry is maturing, and with maturity comes the need for reliability over novelty. Luna: What about the talent shortage? Can the software layer solve the labor crisis if we don't have enough skilled programmers? Lucas: This is the core promise of the new middleware. Natural language interfaces. You point at a task, say 'pick and place,' and the system figures out the kinematics. It lowers the skill floor dramatically. Luna: If that works, it completely changes the economics of automation for small and medium enterprises. Lucas: It does. Because currently, only the giants can afford the teams to customize these systems. Democratizing the software could unlock the mid-market in a way we haven't seen yet. Luna: Do you think the hardware companies will allow that? Or will they try to lock users into their own ecosystems? Lucas: They will try to lock users in, but the market pressure is too strong. If a competitor offers a robot that is ten percent cheaper and twice as easy to program, the buyer will switch. Economics drives interoperability eventually. Luna: So the battle isn't just about better robots anymore. It is about who gives the customer the least headache on Monday morning. Lucas: Precisely. The value proposition is shifting from 'look how smart my robot is' to 'look how little work it takes to keep it running.' That is a much harder sell, but a much more sustainable business. Luna: It sounds like the era of the hero robot is over. The era of the reliable worker is beginning. Lucas: I like that framing. Reliable over heroic. It is a subtle shift, but it defines the next decade of industrial technology. Luna: And for the listeners out there managing these deployments, what is the one thing you would advise them to prioritize right now? Lucas: Prioritize the integration plan before you buy the hardware. Ask the vendor about their middleware capabilities, their API stability, and their support for third-party tools. Do not fall in love with the arm; fall in love with the connection. Luna: Practical advice. It is easy to get distracted by the shiny metal when the real work happens in the code. Lucas: It is. And as we look ahead to the rest of twenty twenty six, I expect we will see more mergers among the software players. The fragmentation cannot last forever. Luna: A cleaner software landscape means faster adoption. It is good for the industry, even if it means fewer choices for the buyer. Lucas: Standardization always reduces choice but increases utility. It is the trade-off we make for efficiency. Luna: Let's hope the next generation of robots is less prone to jamming and more prone to making our lives easier. Lucas: Here is to smoother workflows and fewer error logs. Thanks for listening.