All 22 Robotics posts we have indexed from venture capital blogs and newsletters, plus the highest-scoring pieces of the past year and the people who write about it most.
Physical AI systems pose novel safety challenges: models fail to recognize harms in physical contexts, humanoids lack certified stopping mechanisms, and insurance frameworks haven't caught up. Meanwhile, AI agents are improvising across internet infrastructure in ways traditional security models can't predict.
Generalization and learned models are making robots economically viable across manufacturing, logistics, healthcare, and specialized domains. The shift from prescriptive to probabilistic systems creates a new category of capability, with deployment becoming crucial to further improvement and the largest effects likely coming from tasks that become possible once physical work is cheap.
Agility Robotics' $1.8M revenue against $140M burn hinges on a $8,500/month RaaS price that achieves cost parity with warehouse labor only at two-shift utilization; the entire business plan depends on manufacturing costs falling from $150K to $30K per unit, requiring $150M of robot inventory built before revenue materializes—solvable via equipment financing, but requiring both capital sophistication Agility hasn't yet deployed and customer renewal discipline it cannot control.
A VC framework decomposing the Physical AI market across horizontal layers (models, data/simulation, inference/ops, full-stack platforms) and twelve verticals, mapping 250+ companies as labor shortages, generalization capabilities, and infrastructure maturity converge to unlock automation of complex real-world tasks.
The robotics industry faces a critical chasm between functional demos and reliable production systems—requiring task-specific data, hardware redesign for durability, and 99.9%+ reliability versus lab-bench performance. Vertical applications are narrowing this gap as deployments generate the data and playbooks to improve the next generation.
As autonomous systems and AI agents make real-time decisions without human intermediaries, traditional insurance underwriting breaks down. A new model—trace-economic underwriting using AI telemetry and parametric triggers—could price individual actions, forcing founders to build systems that are fundamentally insurable.
As robots move from prototype to production, builders need innovation across three layers: data and simulation (closing robotics' 100,000-year data gap), hardware and networking (real-time teleoperation, onshoring, better actuators and sensors), and software tooling (MLOps for robots, fleet orchestration, guardrails).
World models—AI systems that understand 3D space, physics, and causality—represent the next multi-trillion-dollar frontier, shifting AI from text to physical labor. The piece analyzes competing architectures (generative vs. predictive latent) and positions NVIDIA, Alphabet, Tesla, Meta, and Apple as dominant players across infrastructure, data, robotics, and edge layers.
Radical VC (4) · Not Boring (3) · Tanay Jaipuria (3) · Amir Kabir (3) · Dynamo VC (1) · Greylock News (1) · Making Connections by Jax (1) · 99% Derisible (1) · Investing in AI (1) · Bilal Zuberi (1)
Suggest a blog or newsletter for the SandHill index. We track the RSS feed and feature the best pieces in the weekly digest.
The latest theses & analyses from 800+ venture investors, curated into one email. Join 5,000+ investors, founders and LPs.