Humanity’s Last Machine is a comprehensive report on humanoid robotics that begins at the component level and expands outward to examine the broader industrial and geopolitical ecosystem shaping the field, tells Hardware FYI. Designed as both a research document and an interactive site, it models what rigorous technical analysis can look like, with a downloadable PDF for offline reading. Authored by investors and operators active in robotics, the report moves systematically from mechanical hardware to sensing, compute, supply chains, and global capital flows. Its most detailed analysis centers on actuation, widely regarded as the primary bottleneck in scaling humanoid production.
Actuators and joint assemblies represent the largest and least scalable cost component in a humanoid’s bill of materials. A typical humanoid robot uses roughly 25–30 actuators, making joint cost a dominant driver of total system price. For context, Unitree’s actuators are estimated to cost around $300 per unit, underscoring how quickly costs compound at scale. Unlike compute, which benefits from semiconductor economies of scale, precision electromechanical assemblies remain difficult to commoditize.
On the compute side, most humanoid platforms rely on Nvidia’s Jetson modules rather than custom silicon. This approach reduces development cost and risk while offering sufficient performance for neural network inference. Performance is measured in TOPS, TOPS per watt, and memory bandwidth. Tesla stands apart with Optimus running on its proprietary Full Self-Driving SoC, reflecting a vertically integrated strategy.
The report also highlights divergent capital strategies between the United States and China. U.S. investment skews toward software, machine learning, and foundational AI models. China focuses on manufacturing scale, leveraging established supply chains from electric vehicles, drones, and consumer electronics. The result is a robotics race shaped not just by algorithms, but by actuators, factories, and geopolitics.