Key Takeaways
- Physical AI software is stuck in the pre-Android era, facing a “50 different operating systems” problem that chokes modern AI deployment on vehicles, according to Applied Intuition co-founder Peter Lewig.
- Building a specialized operating system (OS) is essential for autonomous machines, focusing deeply on safety-critical real-time control, precise memory management, and highly dependable software updates.
- Applied Intuition's $15B valuation comes partly from this approach: creating a universal OS that supports diverse chipsets, letting customers swap out autonomy tech while ensuring strong core performance.
- Unlike typical vehicle manufacturers who “never” update their software, a dedicated physical AI OS allows for frequent, dependable over-the-air updates, a capability Peter Lewig compares favorably to Tesla's monthly releases.
The "50 OS" Problem Choking Physical AI
Imagine Google trying to launch an app in the early days of mobile, only to find phones running 50 different, incompatible operating systems. That's the messy reality Peter Lewig, co-founder of Applied Intuition, says today's physical AI developers face. Lewig describes the current state of physical machine software as “more akin to the state of the phone market before Android and iOS existed.” He points to Google's Android origin story: Larry Page bought dozens of phones, found them all running unique OSs, and realized they needed a consolidated platform to run their products reliably.
Today, autonomous vehicles and other physical AI applications struggle with the same fragmentation. Legacy systems, often from different vendors or designed for narrow purposes, mean that getting “a modern AI application to run on these vehicles, you actually you first have to consolidate the operating system.” This isn't just an inconvenience; it's a fundamental roadblock to progress. Without a shared, stable foundation, deploying sophisticated AI becomes an exercise in endless, bespoke integration—a slow, expensive, and error-prone endeavor.
Why a Specialized OS Isn't Optional for Safety
Consolidating the OS isn't just about convenience; it's about the very core of safe, dependable operation in the physical world. For AI in vehicles, an OS needs to go “deep into the safety critical realm and embedded systems,” Lewig explains. This means managing the “real-time control of let's say the electric motors or the the engine and the actuators” and handling complex redundancies for systems like steering. A general-purpose OS just doesn't cut it when human lives are on the line and milliseconds matter.
Beyond immediate control, there's another often-overlooked but equally critical aspect: updates. Lewig notes that while his Tesla gets updates “fairly frequently, right, once a month,” most vehicle manufacturers “are basically never doing updates.” Even when they do, it's often for just one isolated module. A physical AI OS, like the one Applied Intuition is building, must enable reliable, system-wide updates. This capability is essential for patching vulnerabilities, improving performance, and deploying new AI models without catastrophic failures, moving physical AI from a static, deploy-and-forget model to a continuously improving, software-defined future.
Applied Intuition's Infrastructure Play
Applied Intuition's strategy isn't about owning the entire stack; it's about providing the foundational layer that makes the rest possible. “Our philosophy is that we are a technology company,” Lewig states. They license their OS, acting as the neutral platform that enables others to build. This means a customer can license Applied Intuition's full autonomy tech stack, or they can simply license the OS and integrate their own AI technology on top. This approach, similar to how Android provides a platform for diverse phone manufacturers and app developers, separates the underlying infrastructure from the application layer.
By offering a universal OS that supports diverse chipsets and comes with strong documentation and developer tooling, Applied Intuition is betting that the winning strategy for hard tech isn't vertical integration from day one, but rather providing the shared, high-quality infrastructure that unlocks innovation across an entire industry. They allow their customers to choose their autonomy tech, preventing vendor lock-in while ensuring a stable, safety-first operating environment.
What to Do With This
If you're building in hard tech, especially around physical AI, stop looking only at your application layer. Pull your systems architecture diagrams. Identify every embedded system and the OS it runs. If you find more than three disparate OS environments managing critical functions, you're building on a mobile phone market pre-Android. Start mapping out your OS consolidation strategy now, before your AI capabilities hit an infrastructure wall.