We previously touched on one of the key characteristics of Vertical Integrators being the combination and integration of cutting-edge, but largely proven hardware technology, robotics, and AI software to drive a meaningful step change in efficiency and scalability in a complex, physical value chain.

Riding these exogenous technology curves is a key driver to outcompete incumbents and maintain relevance or even dominance for decades, rather than becoming the sluggish incumbent themselves.

The improvements in foundational hardware technologies and the commoditisation of SaaS/AI create the perfect storm for a new wave of cost-effective, yet powerful physical end-products.

Hardware Technology Drivers

Over the past decade, a set of once-specialised hardware technologies has become cheap, modular and globally standardised, reshaping the economics of building physical products. What used to require large engineering teams and bespoke supply chains can now be assembled by small, lean startups, thanks to cost curves that continue to fall and performance that keeps compounding.

Battery technology is the clearest case: Pack prices for lithium-ion have fallen nearly 90% since 2010, averaging ca. $115/kWh in 2024, with LFP cells in China reported at $50-60/kWh. New chemistries, such as LMFP cathodes and sodium-ion designs, are reaching unparalleled energy density of 180 Wh/kg at cell level, while early solid-state or lithium-metal prototypes exceed 350 Wh/kg at the cell level. Over the next few years, mainstream batteries are expected to gain another 10-20% in energy density, with premium solid-state formats achieving 25-30% or more in niche applications. For mobile robots, drones, and wearables, that translates directly into lighter systems or longer endurance without redesigning the rest of the platform.

Edge compute has undergone a similar transformation: Edge-AI modules such as NVIDIA’s Orin Nano already offer 60-100 TOPS of inference for under $500, and emerging low-power application-specific integrated circuits (ASICs) are bringing high-performance vision and planning to devices running on single-digit watts. This allows real-time autonomy without an always-on cloud link, reducing latency, bandwidth costs, power requirements and privacy risk, while increasing reliability and unlocking new use cases.

Sensing has seen not just falling prices but leaps in quality, size, and robustness. Automotive lidar has dropped from $75k prototypes to $300-500 production modules, headed below $200 for ADAS. mmWave radar chips cost about $10-15 in volume, and 6-9-axis IMUs are only a few dollars. Beyond cost, sensors are smaller, better sealed, and more tolerant of shock, vibration, and weather - key for reliable operation in harsh and unpredictable environments. That enables rugged and powerful multi-sensor suites across vision, inertial, and environmental, on assets as diverse as drones, farm equipment, and industrial robots, without forcing compromises in weight or reliability.

Connectivity is now largely plug-and-play: NB-IoT and LoRa modules cost just a few dollars, and LTE-M, BLE, and Wi-Fi come with integrated stacks and secure over-the-air updates. Prototyping and manufacturing tools/providers are a key enabler - two-layer PCBs from major fabs cost only a few dollars with 24-hour turnaround, while desktop CNC and additive tools let startups iterate in hardware almost as quickly as in software.

Even access to orbit has normalised, with Falcon-class rockets cutting launch costs to ≈$2k/kg, opening constellations and space-to-ground services to small companies - a further step-change is expected once SpaceX’s Starship becomes operational and other micro-launchers start to scale. Starlink and other low-earth-orbit (LEO) constellations have added a global, high-bandwidth layer. Hardware teams can deploy connected assets in remote environments, oceans, or disaster zones without building bespoke telecom infrastructure, while also streaming data for testing and updates anywhere on the planet.

Lastly, the “middle layer” that binds disparate hardware components into coherent systems has been modernised. Standardised middleware - ranging from ROS 2 in robotics to lightweight real-time operating systems, modular device drivers, and protocol-translation frameworks - is making batteries, sensors, compute boards, and communications modules far more interoperable, by abstracting away their quirks while bundling safety, certification, and security primitives. Modern stacks now ship with plug-and-play APIs, auto-discovery of peripherals, and cloud-native management services, reducing the custom firmware once needed to make components talk to each other.

This maturation of the integration layer means teams can focus on architecture and product logic instead of low-level plumbing, accelerating time-to-market and allowing small startups to assemble sophisticated hardware-software stacks with a fraction of the historical engineering overhead.

Together, these curves push the centre of gravity in terms of value capture away from component invention toward integration, software, certification, and operations. Winning startups assemble cheap energy storage, efficient compute, high-fidelity sensors, ubiquitous connectivity, and rapid fabrication into reliable, certifiable systems. Then ride exogenous technology curves for compounding gains. A drone designed today might fly 15-20% farther in three years simply by adopting a new pack chemistry, ceteris paribus, while better sensors and lighter compute boards will expand its capability envelope without redesigning the airframe. 

For vertical integrators, this dynamic compounds: tighter feedback loops enables rapid iteration between technology, manufacturing, and field data to help them improve capability and scale unique data assets to out-execute incumbents on both time-to-market and unit economics.