A new generation of consumer hardware is emerging—one that doesn’t just capture images, but interprets the world in real time. Switzerland-based Mosaic SoC is positioning itself at the center of that shift, announcing a $3.8 million pre-seed round to develop specialized chips designed for spatial awareness in power-constrained devices.
The round was led by Founderful , with participation from Kick Foundation .
The Bottleneck Behind Spatial Computing
Devices today are equipped with increasingly sophisticated sensors and cameras, but interpreting that data remains resource-intensive. Real-time perception—understanding position, movement, and surroundings—typically relies on power-hungry processors or GPUs, limiting its use in smaller, battery-sensitive devices.
This constraint has slowed progress in areas like augmented reality glasses and always-on mobile vision systems. While prototypes exist, scaling these capabilities into everyday consumer products has proven difficult due to trade-offs in heat, battery life, and device size.
Mosaic SoC is attempting to address that bottleneck with a different approach: a dedicated perception chip that handles spatial intelligence independently, reducing reliance on larger processing stacks.
A Dedicated Layer for Real-Time Perception
Rather than adding more general-purpose compute, Mosaic SoC’s architecture is designed specifically for processing visual and positional sensor data. The company’s chips aim to give devices a continuous understanding of their surroundings while operating within tight energy budgets.
In practical terms, this enables devices to build local maps of their environment, recognize objects, and track movement without constant cloud processing or heavy onboard computation. For wearables, this could mean AR glasses that maintain awareness throughout the day without noticeable battery drain. For smartphones, it introduces the possibility of always-on computer vision features that activate only when relevant events occur.
Designed for Integration, Not Complexity
A recurring challenge for hardware manufacturers is the complexity of integrating new capabilities into existing systems. Mosaic SoC’s model attempts to reduce that friction by bundling its chips with a full application layer.
This means original design manufacturers (ODMs) can integrate spatial intelligence without building the software stack from scratch. Instead of adding engineering overhead, the chip is positioned as a foundational layer that developers can build on.
The company generates early revenue through non-recurring engineering (NRE) contracts with partners, with expectations of shifting toward scalable chip sales as production ramps.
Founders Focused on Edge Intelligence
Mosaic SoC was founded by Moritz Scherer and Alfio Di Mauro, both PhDs from ETH Zurich with backgrounds in system-on-chip design.
Their focus is on closing the gap between the growing demand for edge-based intelligence and the limitations of current hardware architectures. Instead of scaling traditional CPU-based designs, the company has developed a proprietary multi-core system optimized for performance per watt, targeting always-on perception use cases.
From Chips to Ecosystems
As specialized perception hardware matures, the center of gravity may shift from raw silicon performance to the software layers that sit on top of it. Toolchains, compilers, and developer frameworks increasingly determine how easily new capabilities can be deployed and how widely they are adopted. In this model, chips become the foundation, but the surrounding ecosystem—how developers build, optimize, and distribute applications—ultimately shapes the pace of innovation.
This transition mirrors a broader industry pattern where hardware alone is no longer the primary differentiator. Instead, the combination of tightly integrated hardware and software environments can create lock-in, standardization, and new development paradigms, particularly in emerging categories like spatial computing.
Toward Persistent Spatial Awareness
If low-power, always-on perception becomes viable at scale, it could fundamentally change how devices interact with users and their surroundings. Rather than responding only when activated, devices could maintain a continuous understanding of context—location, movement, objects, and intent—without relying heavily on cloud processing.
That shift introduces new possibilities, but also new considerations. Persistent awareness could enable more intuitive interfaces and automation, while raising questions around privacy, data ownership, and how much real-time interpretation should occur on-device versus externally. It may also blur the line between passive tools and active systems that anticipate user needs.
The technical challenge is no longer just achieving real-time perception, but doing so efficiently, securely, and in a way that integrates seamlessly into everyday devices. Whether this category becomes mainstream will depend as much on those trade-offs as on the underlying performance gains.