Software becomes the primary robotics bottleneck
The shift from hardware limits to a robotics software bottleneck describes how performance, safety, and innovation are now constrained more by software architecture, integration, and security than by mechanical or electronic components, especially as physical AI development moves into shared human environments and must meet tighter reliability and certification demands. QNX’s Inside the Robot report, based on 1,000 robotics developers, shows almost one in three respondents (27%) now see software architecture and integration as their biggest performance bottleneck, compared with 16% who blame hardware. This marks a clear break from the era when faster processors or better sensors were the main constraint. As systems grow more AI-enabled and interconnected, developers say future progress depends on building predictable, secure platforms that can handle mixed‑criticality workloads. The study also finds 85% of respondents expect software’s role in robotics to grow further over the next three to five years.
Physical AI leaves the lab and enters human environments
The QNX research shows physical AI development is rapidly shifting from fenced-off pilot lines to unconstrained, human-centered spaces such as shop floors, surgical suites, and city streets. More than four in five respondents (83%) report their robots are already deployed alongside people, and two‑thirds of the remaining group expect this to happen within three to five years. This move raises the stakes for reliability, deterministic timing, and predictable behavior. Nearly all surveyed developers (95%) say deterministic, real-time execution is important to their systems, yet 91% still run at least part of these workloads on general-purpose operating systems. This gap between requirements and tooling increases functional safety risks and slows certification, especially in settings where human-machine interaction is constant. The result is a growing tension: teams value the flexibility of familiar software stacks, but are pushed toward safety-certified platforms to meet rising expectations around robotics security challenges and compliance.
Performance, security, and scalability limit physical AI
QNX’s benchmark report highlights four recurring pain points for developers: integration complexity, certification delays, functional safety risks, and unpredictable behavior under load. These issues cut across performance, security, and software scalability in robotics, and they are slowing down physical AI even when advanced hardware is available. As robots take on more autonomy, they must combine perception, planning, and control in one system while handling mixed levels of criticality. At the same time, cybersecurity has moved to the center of robotics security challenges. According to QNX, 51% of surveyed teams expect their largest near-term investments to go into AI-driven decision making and another 51% into cybersecurity. These priorities reflect the reality that connected robots are potential attack surfaces as well as safety-critical assets. Without strong isolation, real-time guarantees, and clear certification paths, adding more compute or more AI models does little to close the gap between ambitious use cases and deployable systems.
Architectures shift toward software-first design
The survey suggests a broad architectural shift as teams respond to the robotics software bottleneck. Developers are rethinking operating system choices, middleware, and partitioning strategies to achieve predictable behavior alongside flexibility. Although 91% still rely on general-purpose operating systems for at least part of their stack, 86% of those users say they are open to changing their OS, a sign that many are reaching the limits of current platforms. New architectures focus on real-time kernels, safety-certified components for critical paths, and clear separation between high-assurance control and higher-level AI experimentation. Operating systems and real-time control software are cited by 37% of respondents as a major investment area, indicating a turn toward software foundations as strategic assets. For developers, this means embracing software-first thinking: designing for certification from day one, planning for distributed deployments, and prioritizing predictable, secure execution as highly as raw AI performance.
